Standardizing Motion Resolution: "Milliseconds of motion resolution" (MPRT) better than "Lines of motion resolution" - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #1 of 72 Old 09-15-2013, 08:21 AM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
History:

[Edit, added to add context]
Reviewers often use "Lines of Motion Resolution", on motion test patterns based on FPD tests:
Search: "1080 lines of motion resolution" -avsforum (excludes AVSFORUM posts)
Search: "1200 lines of motion resolution" -avsforum (excludes AVSFORUM posts)

Examples:
- C|Net Review of HDTV measuring motion blur by "lines of motion resolution"
- HDGuru talks about motion blur by "lines of motion resolution"
- HDTV Magazine: Panasonic Viera touting "lines of motion resolution"
And many home theater magazines sitting here, using "lines of motion resolution" in their HDTV reviews.

Details:

Lines of motion resolution is very test-pattern specific.
- depends on speed of motion
- depends on test pattern
- depends on resolution of display (1080p versus 4K)

Milliseconds of motion resolution is test-pattern independent.
- can be measured by scientific equipment
- can be measured by human eye with certain special test patterns
- resolution independent, refresh rate independent.

Google: Moving Picture Response Time
  1. It's resolution independent.
  2. It's motion speed independent. (most displays have the same MPRT at all motion speeds and motion vectors)
  3. It's more test pattern independent.
  4. It's future proof. 1080p, 4K, 8K, VR, etc.
  5. Improvement is unbounded. It doesn't cap out at a specific value (e.g. "1200 lines of motion resolution")
  6. It easily covers the faster motion speeds often seen in video game use, an increasing use case of displays.

Easy consumer motion test patterns needs to be developed over the coming years, to discontinue the "lines of motion resolution" lingo. Even disregarding the subjectiveness problem (like Contrast), "lines of motion resolution" just simply has a far bigger superset of problems far bigger than "milliseconds of motion resolution" or "MPRT" or "moving picture response time" (which is NOT the same measurement standard as "LCD response time"). It may take time (e.g. Blu Ray becoming obsoelte), but it's time to begin migrating to a a future proof motion resolution standard.

___________________________________________________________________

I'm the author of the Blur Busters UFO Motion Tests.

I feel that motion resolution needs to be I feel it's better for motion resolution to be measured in milliseconds (MPRT), rather than motion resolution measured in "lines of motion resolution". MPRT, stands for Motion Picture Response Time. Most reviews quote a "lines of motion resolution".

With MPRT, you know that mathematically, 1 millisecond of motion blur equals 1 pixel of motion blur for every 1000 pixels/second motion. Very simple math. e.g. 4ms MPRT means you get 8 pixels of motion blur during 2000 pixels/sec motion. Apples to apples, even comparing 4K displays to 1080p displays. Note that MPRT is a different measurement than LCD pixel transitions (e.g. GtG transitions), the closest analog is the time it takes for a pixel to go BWB (black-white-black).

It would be good to see more Blu-Ray motion resolution tests to migrate to the MPRT standard, rather than "Lines of Motion Resolution". Or alternatively, motion equivalence ratios (1000 / MPRT). So a display with a true "250" measured motion equivalence ratio has about 4ms of MPRT. (1000 / 4) = 250. True measured motion equivalence numbers such as "250" is also the same amount of motion blur as a hypothetical "250fps on a perfect 250Hz sample-and-hold display", and may be more user-friendly than millisecond numbers.

That said, it's easy to convert motion equivalence ratios mathematically as 1000 / MPRT = motion equivalence ratio (MER). This is measured analog to the claimed numbers often quoted on displays (e.g. Samsung CMR 960, Motionflow XR 480, Panasonic 1600 scanning backlight) and the measured motion equivalence ratios would certainly fall short of these numbers, much like measured contrast ratios fall short of claimed numbers.

Quoting by MPRT or by MER is more apples to apples.

- Resolution independent
- Test pattern independent
- Can be measured by eye or measured by scientfic equipment
- Easy to extrapolate

With the transition from 1080p to 4K, it's time to migrate away from the archaic "Lines of motion resolution" standard, and go to the modern "milliseconds of motion resolution" (MPRT) or its simple inverse, "measured motion equivalence ratio" (MER)

Edit as of June 2014: Industry has started talking about persistence more often recently. "Milliseconds of persistence" is another way to describe "milliseconds of motion resolution", since it's the same number..

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
Sponsored Links
Advertisement
 
post #2 of 72 Old 09-15-2013, 01:40 PM
AVS Special Member
 
Doug Blackburn's Avatar
 
Join Date: May 2008
Location: San Francisco - East Bay area
Posts: 3,453
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 19 Post(s)
Liked: 226
Maybe it's better... have to think about it a bit more, but like every attempt to quatify motion blur, it ignores one significant factor... blur is not a constant. Luminance affects blur and the associated colors affect blur. For example, if you have a magenta object moving on a green background, you may have a lot of blur but when the magenta object moves over a gray or white or black or orange background, blur results could be very different. So any one number defining motion blur is only good for one specific test condition and that one test condition cannot possibly define the video display's overall blur vs. blur-free performance. And do we really care if one display is 2% better than another? No. What we care about is that LCD displays dip down to 200-300 pixels (or equivalent) of resolution during motion versus plasma or DLP displays that can maintain 900 pixels of resolution or more during motion. Does it really matter if one LCD is the equivalent of 280 pixels while another one can do the equivalent of 310 pixels? Not really. What might be important is getting LCD up to the equivalent of 600 pixels of resolution during motion... and that's easy enough to see if you have used something like DisplayMate's motion quality evaluation patterns and images. Their patterns include a wide range of luminance values, plus B & W and color and even the color is represented by a range of luminances. Once you've seen half a dozen LCD displays reproduce those patterns, if you see a plasma or DLP or a particularly "fast" LCD, you can tell just from those patterns that the result is better. Do we care what the number is? Eh... if it ever becomes an "official" spec, it will be abused just as badly as contrast ratio or projector lumens or the gain of acoustically transparent screens or amplifier watts in "real" amps vs. AVRs, or just about any other over-stated spec we all wish was more meaningful.

"Movies is magic..." Van Dyke Parks
THX Certified Professional Video Calibration
ISF -- HAA -- www.dBtheatrical.com
Widescreen Review -- Home Theater & Sound
Doug Blackburn is offline  
post #3 of 72 Old 09-15-2013, 05:19 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post


With MPRT, you know that mathematically, 1 millisecond of motion blur equals 1 pixel of motion blur for every 1000 pixels/second motion. Very simple math. e.g. 4ms MPRT means you get 8 pixels of motion blur during 2000 pixels/sec motion.

Is the point of your post that people should pay attention more to MPRT rather than "lines of motion resolution"?

Also, while the MPRT tells you about motion blur, it says nothing about eye tracking based motion blur right?
spacediver is online now  
post #4 of 72 Old 09-15-2013, 05:22 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Doug Blackburn View Post

Maybe it's better... have to think about it a bit more, but like every attempt to quatify motion blur, it ignores one significant factor... blur is not a constant. Luminance affects blur and the associated colors affect blur. For example, if you have a magenta object moving on a green background, you may have a lot of blur but when the magenta object moves over a gray or white or black or orange background, blur results could be very different. So any one number defining motion blur is only good for one specific test condition and that one test condition cannot possibly define the video display's overall blur vs. blur-free performance.

Will different displays be affected differentially with respect to these different conditions? Or will they all show the same amount of relative motion blur in these different conditions?

So, for example, take two displays, A and B. Are you saying that A's motion blur performance might be affected by luminance more so than B's motion blur performance?
spacediver is online now  
post #5 of 72 Old 09-16-2013, 12:38 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by Doug Blackburn View Post

Maybe it's better... have to think about it a bit more, but like every attempt to quatify motion blur, it ignores one significant factor... blur is not a constant.
Correct, sir.
--> LCD black-to-white transitions can be faster than LCD white-to-black transitions, and various other shades
--> Plasma motion resolution for dark murky shades versus plasma motion resolution for primary colors
--> PWM dimming can increase slightly motion resolution, because of the accidental black frame insertion effect between the last pulse of the previous frame, and first pulse of next frame. LCD's can vary in MPRT by a few milliseconds (e.g. 14ms vs 16ms) when dimmed, as a result.
--> Other colors can be different in speed in transitions.

HOWEVER, it equally affects "lines of motion resolution" measurements as well as "milliseconds of motion resolution" measurements equally, for the same test pattern case. The bottom line point STILL stands: "line of motion resolution" is test pattern specific. "milliseconds of motion resolution" is test pattern independent.

However, there's solutions. It is possible to design special motion tests that uses a matrix of variety of colors, and benchmark the average motion resolution of a set of about one-dozen color transitions, to come up with an excellent average motion resolution. I also invented a new inexpensive pursuit camera technique, which should be studied by display reviewers.
Quote:
Luminance affects blur and the associated colors affect blur. For example, if you have a magenta object moving on a green background, you may have a lot of blur but when the magenta object moves over a gray or white or black or orange background, blur results could be very different. So any one number defining motion blur is only good for one specific test condition and that one test condition cannot possibly define the video display's overall blur vs. blur-free performance.
Correct.
Quote:
And do we really care if one display is 2% better than another?
The differences are not that small.
We are talking about order-of-magnitude differences in motion clarity:



Yes, a full order of magnitude scientifically measured difference -- 16.7ms versus 1.4ms
And it actually even corresponded with visual observations (motion blur trail length was subjectively one-tenth as much -- e.g. 36 pixels of motion blurring became only 3 pixels of motion blurring). Remember, blur in computer graphics are more demanding than video/movies.

In 2012 and 2013 there were quite interesting developments in LCD that allowed startling increases in motion clarity in LCD -- please see TFTCentral's article (TFTCentral: Motion Blur Reduction Backlights. Unlike yesterday's scanning backlights that only improved motion clarity by a few percent, newer ultra-high efficiency full-strobe backlight technologies manage to reach the theoretical motion-blur-reduction abilities linearly with strobe length (halved strobe length = exactly double motion resolution).

During my testing with LightBoost, this is a witnessing of an LCD panel that has 12 times the motion clarity (e.g. 3 pixels of motion blur on one panel that's 2ms speed, and on a different LCD, over 36 pixels of motion blur on a different panel that's also 2ms in speed). This would be equivalent to 3,600 lines of motion resolution for a test pattern that topped out at 300 lines of motion resolution, but the motion test patterns stop at not much beyond 1000 lines of motion resolution, so newer tests are needed to see well beyond this.

Videogame motion is far faster than movie motion, and television motion, and videogame graphics are sharper. Tests have shown that there are very noticeable motion clarity improvements between a 2.4ms strobe length and a 1.4ms strobe length (which is akin to comparing a 400fps@400Hz display, versus a 700fps@700Hz display) -- because it added about 2 pixels of motion blur to the very-very tough Panning Map Test at http://www.testufo.com/#test=photo&photo=toronto-map.png&pps=1440 (very few displays pass the map-label readability test at 960 pixels per second -- less 1440 pixels per second -- namely CRT displays, certain Panasonic NeoPDP's, and certain LightBoost LCD displays). This panning map is completely unreadable on a 60Hz computer monitor. But amazingly, the panning map is perfectly readable on a LightBoost LCD (see testimonials) which is one of the first LCD to actually have less motion blur than CRT's (at least medium-persistence CRT's).

So a display with a theoretical 2.4ms strobe length (which would translate to 1,800 lines of motion resolution for a test pattern at 300 lines @ 60Hz) and a 1.4ms strobe length (which would translate to 3,600 lines of motion resolution for a test pattern at 300 lines @ 60Hz).

We're not talking about 2 percent differences -- we're talking about major differences -- and specific use cases like panning, turning, strafing in video games, and computer graphics.

Did you know that Eizo has just released a new 240Hz strobe-backlight LCD computer monitor, designed to reduce motion blur eyestrain during panning satellite maps?
http://www.eizo.com/global/products/duravision/fdf2405w/

How do we test this monitor? And such future monitors? Blu-Ray motion resolution test patterns definitely would not cut it

List of Interpolation-Free Ultrahigh-Efficiency Strobe Backlights
(strobe backlights that reach near scientific theoretical motion-blur-reducing efficiency; according to motion-blur-trail-length measurements of fast-moving photo tests)
-- Sony Motionflow Impulse (HDTV)
-- Eizo's FDF2405W (monitor)
-- nVidia LightBoost (monitor)
-- Samsung 120Hz 3D Mode (monitor)

These ultra-highefficiency strobe backlights only came out in the last 2 years; and is a new area of LCD technology, that is producing dramatic jumps upwards in motion resolution in the last two years alone on certain LCD's, and my measurements have confirmed the order-of-magnitude chasm between the worse LCD and the best LCD. Today's motion resolution test patterns are no longer suitable to cover this whole motion blur gamut. It's time to move to "milliseconds of motion blur" to allow this progress..

Provisionally, it's possible to convert certain test patterns into MPRT, for a specific test pattern calibrated to result in exactly 300 lines of motion resolution for a 60Hz sample-and-hold display (60Hz).
300 lines = 16.7ms (1/60sec)
600 lines = 8.3ms (1/120sec)
1200 lines = 4.1ms (1/240sec)

And beyond limits of blu ray test patterns
...2400 lines (extrapolated) = 2.08ms (1/480sec)
...4800 lines (extrapolated) = 1.04ms (1/960sec)

Also, new displays such as OLED can be pulse-driven.
OLED's with a 8.3ms pulse-drive would yield 600 lines of motion resolution (or an 8.3ms MPRT). OLED pulse lengths will shorten dramatically over the next 5 years, causing MAJOR swings in motion resolution.

Video and movies can afford add source-based motion blur. But a lot of us videogame-players hate the videogame adding artifical motion blur at the source, and thus, the graphics are unadulterated by source-based motion blur, and as we approach 4K, the sample-and-hold effect becomes a major bottleneck.

Virtual reality googles are an excellent extreme test case of motion blur. Turning your head slowly at 30 degrees per second yields about 1920 pixels per second of motion on a 1080p VR screen enveloping your vision. That creates 4 pixels of motion blur at 2ms strobe lengths (blur equivalence to 500fps@500Hz) and creates 2 pixels of motion blur at 1ms strobe lengths (blur equivalence to 1000fps@1000Hz).

John Carmack, Michael Abrash, and Oculus VR agrees with this
This is confirmed by vision resarchers, who have come up with impressive research in recent years that the final frontier is not remotely 60Hz or 120Hz. The point of diminishing returns do not end for a very, very, very, very long time well beyond 120Hz. Human vision can already tell apart versus in virtual reality tests, confirmed by myself, confirmed by Michael Abrash of Valve Software (link), confirmed by John Carmack (his QuakeCon keynote about motion blur) who now work at Oculus. (BTW -- I stay in contact with both of them; and they are fans of Blur Busters Blog, as well as having contacted Palmer Luckey of Oculus). You will also notice Michael Abrash also wants to see a theoretical 1000Hz computer display someday; and I agree with his sentiments.

So, Blu-Ray Test Pattern Creators (all of you), please stop using "Lines of motion resolution". It's a stupid measurement that's not future-proof. Begin using the properly scientific "Milliseconds of motion resolution" (or its inverse, a motion equivalence ratio, as 1 / MPRT). Vision resarchers prefer that.

To demonstrates limitations of today's Blu-Ray motion resolution test patterns, I fully challenge you to try to read the map labels* on this TestUFO Moving Map Test at 1440pixels/second Many displays that succed "1080 lines of motion resolution" fail this Panning Map Test. Only displays that exceed about ~2500 lines of motion resolution (extrapolated, since blu-rays don't measure that far) allow you to be able to reliably read the street name labels on this Panning Map test. A very few plasmas (e.g. NeoPDP) and a rareified very, very, very few LCD's (e.g. LightBoost), succeed this test. And this Panning Map Test -- panning speed is STILL slower than a 30-degree-per-second head turning during Virtual Reality Goggles!

*IMPORTANT: View TestUFO links in a supported web browser capable of perfectly framerate-refreshrate synchronized animations (you need a fast computer with a recent nVidia/AMD GPU).

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #6 of 72 Old 09-16-2013, 01:06 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

Is the point of your post that people should pay attention more to MPRT rather than "lines of motion resolution"?
Not necessarily. See my reply above.
Quote:
Also, while the MPRT tells you about motion blur, it says nothing about eye tracking based motion blur right?
MPRT includes everything -- including tracking-based motion blur.
That means the 60Hz LCD monitors that are 2ms in transition time, are actually 16.7ms in MPRT rating precisely because of eye-tracking-based motion blur. Sample-and-hold motion blur is exactly the same thing as eye-tracking-based motion blur. So the manufacturer transition ratings are quite meaningless because they are lost in the noisefloor of the motion blur caused by high MPRT's caused by tracking based motion blur.

(For those who do not understand sample-and-hold motion blur -- see Animation of Eye-Tracking Motion Blur -- www.testufo.com/#test=eyetracking) ... Tracking based motion blur is always part of a moving test pattern & pursuit camera.

As you can see in my previous post, "Lines of motion resolution" also accounts for eye tracking motion blur, and this is why "Lines of motion resolution" is convertible to "MPRT" but only and only for a specific test pattern at a specific test speed, while MPRT measurements are more test-pattern independent (especially if you standardize on averaing the MPRT for all color transitions). e.g. 300 lines of motion resolution @ 60Hz sample-and-hold, would be equivalent to 16.7ms MPRT (1/60 = 16.7)

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #7 of 72 Old 09-16-2013, 01:16 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

Will different displays be affected differentially with respect to these different conditions? Or will they all show the same amount of relative motion blur in these different conditions?

So, for example, take two displays, A and B. Are you saying that A's motion blur performance might be affected by luminance more so than B's motion blur performance?
Yes, brightness can affect motion blur performance,
BUT it is a problem that equally affects both "lines of motion resolution" AND "milliseconds of motion resolution"

The POINT is that "milliseconds of motion resolution" is more universal, more fair, and more apples-versus-apples, than "lines of motion resolution" which has a superset of bigger problems/limitations, especially as we migrate towards future displays and future use cases (more pixels, clearer graphics, lack of source-based blur, faster motion, etc) that currently creates visible limitations that far exceeds the top resolution limit of the current blu-ray motion resolution tests.

There ARE workarounds, such as doing several passes of motion resolution tests for several common colors and luminances, and creating a ever-more-accurate MPRT average that's more representative of mixed-color picture material (movie material, videogame material, computer material, video material).

It's NO LESS legitimate than attempting an accurate ANSI contrast ratios. True contrast ratios have so many variables too (e.g. ambient light, reflections back to screen, backlight bleed, local dimming halos, electron gun beam halos, etc). Creating a motion resolution average (whether it be the test-pattern specific "lines of motion resolution" versus the test-pattern-independent "milliseconds of motion resolution) is no less illegitimate than attempting to measure a single contrast ratio for a single kind of standardized test pattern. Due to things like haloing, checkerboard contrast ratio can even vary depending on how big the checkerboard squares are! So for MPRT tests to be more consistent, this may lead, someday, to an easy, simplified standardized full-color MPRT measurement test patterns that's display technology independent, while also being compatible with electronic measuring equipment. Innovation is necessary to go beyond the limiting and archiac "lines of motion resolution" which is NOT future-friendly.

MPRT is not my invention though... Many scientists have used that in the past.
See scientific references -- papers back to as old as 2001 have used MPRT as a method of measuring motion blur. This was then long forgotten, but surprisingly, this turns out to be the world's fairest method of measuring motion blur. Time to choose it as a standard...

Of course, maybe the general population prefers "Motion Equivalence Ratios". It's easier to compare a display with a "Measured Motion Eqivalence Ratio of 500" (2ms MPRT) versus a display with a "Measured Motion Equivalence Ratio of 330" (3ms MPRT). Equivalence ratios are just a a simple inverse of MPRT. Equivalence ratio = (1000 / MPRT) which is actually a Blur Busters suggestion, but there's laready precendent TV makers already do it (e.g. Samsung's claimed "Clear Motion Ratio 960" and Sony's claimed "Motionflow XR 960". The claimed equivalence numbers, 960, actually measures far more closely to apporoximately an equivalence ratio of 300 or 400. At least until recently, in the interpolation-free modes of the high-efficiency full-strobe backlights -- that are better than scanning backlights -- which actually makes measured values reach closer to theoretical/exaggerated values).

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #8 of 72 Old 09-16-2013, 01:22 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post

MPRT includes everything -- including tracking-based motion blur.


ah right, I missed this part :
Quote:
Note that MPRT is a different measurement than LCD pixel transitions (e.g. GtG transitions), the closest analog is the time it takes for a pixel to go BWB (black-white-black).
spacediver is online now  
post #9 of 72 Old 09-16-2013, 01:26 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post


However, there's solutions. It is possible to design special motion tests that uses a matrix of variety of colors, and benchmark the average motion resolution of a set of about one-dozen color transitions, to come up with an excellent average motion resolution.

That's an excellent idea.
Quote:
Originally Posted by Mark Rejhon View Post

I also invented a new inexpensive pursuit camera technique, which should be studied by display reviewers.

I remember being baffled at this illusion - I couldn't understand how the image changed depending on whether I tracked it with my eyes vs stationary fixation. Took me a while to figure it out - ingenious smile.gif
spacediver is online now  
post #10 of 72 Old 09-16-2013, 01:32 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Examples of why MPRT will matter in future.

Hypothetical Future Displays

2015 hypothetical OLED with 2ms frame samples (either by pulsewidth and/or by interpolation)
...... MPRT of 2ms
...... Motion Equivalence Ratio of 500.
...... same blur as a theoretical scientifically-perfect 500fps@500Hz sample-and-hold display (even if not running at true 500Hz)
...... Desktop Monitor that creates ~4 pixels of motion blur at 1920 pixels/sec motion
...... Virtual reality goggles that creates ~4 pixels of motion blur during 30 degrees per second head-turning motion (assuming 1080p VR panel covering 30 degrees FOV)

2017 hypothetical Strobe-backlight LCD with 1ms frame samples
...... MPRT of 1ms
...... Motion Equivalence Ratio of 1000.
...... same blur as a theoretical scientifically-perfect 1000fps@1000Hz sample-and-hold display (even if not running at true 1000Hz)
...... Desktop Monitor that creates ~2 pixels of motion blur at 1920 pixels/sec motion
...... Virtual reality goggles that creates ~2 pixels of motion blur during 30 degrees per second head-turning motion (assuming 1080p VR panel covering 30 degrees FOV)

2018 hypothetical OLED with 0.5ms frame samples (either by pulsewidth and/or by interpolation)
...... MPRT of 0.5ms
...... Motion Equivalence Ratio of 2000
...... same blur as a theoretical scientifically-perfect 2000fps@2000Hz sample-and-hold display (even if not running at true 2000Hz)
...... Desktop Monitor that creates ~1 pixels of motion blur at 1920 pixels/sec motion
...... Virtual reality goggles that creates ~1 pixels of motion blur during 30 degrees per second head-turning motion (assuming 1080p VR panel covering 30 degrees FOV)*

*30 degrees FOV is being generously narrow for VR goggles. The field of view can be much wider, which produces more opportunity for motion blur to be noticed at the same head turning speed, requiring more pixels or shorter persistence to reduce motion blur even further. Yes, it's amazing that the human eye can still see half-millisecond differences in motion blur (provided the veil of sample-and-hold is not bottlenecking your ability to see this). Again, point of diminishing returns gradually disappear, but do not disappear for a long time.
  • Mathematically, 1ms equals 1 pixel of motion blur for 1000 pixels/sec motion.
  • And measured motion equivalence ratios are very easy to calculate -- 1000 / MPRT
  • And measured motion equivalence ratios are conveniently directly comparable to a theoretically 'perfect' X fps @ X Hz display
  • See? Beautifully simple math.

Yes, reality is different. Due to inefficiencies, measured equivalences are often lower than claimed equivalences. That said, in modern displays, reality actually have shown gradually improving accuracy between scientific measurements and what was visually perceived in motion tests. We also acknowledge future displays are getting "cleaner and cleaner" in motion. The motion blur trail is becoming simpler and less complex than it used to be to the human eye (aka less artifacty) -- as there's fewer and fewer display limitations to interfere (e.g. LCD ghosting that interferes over multiple refreshes). OLED's and high-efficiency strobe backlights have none of the inefficiencies of subfields, temporal dithering, and phosphor decay. Arguably, MPRT's of OLED's and strobe backlights, are more consistent and accurate to the display and science, than ANSI contrast-ratio measurements are to the display technology. I fail to see why MPRT measurements are any less legitimiate than say, ANSI checkerboard contrast ratio (the same contrast ratio can look a little different on different display technologies, but we live with it anyway). What is needed, is a migration to test patterns that easily gives MPRT values or motion equivalence ratio values, in a display-technology-independent manner.

Eye tracking inaccuracies start to become the limiting factor, and different humans don't track motion as well as others. However, extreme cases such as close-view distances (monitors & VR), blur-free source (computers & gaming), ultra resolution (4K), these use cases are rather extreme and show the definitive need for humankind eventual progress beyond "3600" lines of motion resolution (1ms MPRT) which is equivalent of going beyond 1000fps@1000Hz (someday). This is necessaary to achieve analog-like Holodeck motion quality where all motion blur is caused by your brain, and not forced upon you by source, or forced upon you by the display. Including eye tracking motion blur forced upon you due to display limitation of discrete refreshes, etc (because human brain does not operaate tha way) -- basically motion blur above-and-beyond your human brain limitations.

Another possible way of saying MPRT is "persistence". "1ms MPRT" or saying "1ms of motion blur" could be taken to mean "1ms of persistence". As the measurements actually is done very similarly. But that doesn't always specifically say motion blur. Either way. All of them is milliseconds, which is the proper future-proof way of measuring motion blur. Several industry people, such as John Carmack, use the word "persistence". (e.g. John Carmack's YouTube keynote at QuakeCon.)

Good thread: Why we need 1000fps@1000Hz this century -- Valve Software (Michael Abrash) comments.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #11 of 72 Old 09-16-2013, 01:48 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
for the 2017, don't you mean 1000 fps@1000hz?
spacediver is online now  
post #12 of 72 Old 09-16-2013, 01:57 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Fixed, that was a typo.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #13 of 72 Old 09-16-2013, 09:28 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

Quote:
Originally Posted by Mark Rejhon 
I also invented a new inexpensive pursuit camera technique, which should be studied by display reviewers.
I remember being baffled at this illusion - I couldn't understand how the image changed depending on whether I tracked it with my eyes vs stationary fixation. Took me a while to figure it out - ingenious smile.gif
Thanks for the compliment!

I've also been complimented by an european scientist, a vision researcher from Switzerland (Marc Repnow) about this technique I developed. One only needs is a $150 camera rail & a regular consumer camera (with adjustable exposure), and you've got more accurate motion blur photo capture capability formerly the territory of >$10,000 equipment. Brings it within blogger territory and magazine reviewer territory.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #14 of 72 Old 09-16-2013, 09:51 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post

TV manufacturers already use motion equivalence ratios ("240", "480")
The MPRT technique simply brings cheap scientific measurability to it. (MEASURED motion equivalence ratios).
Much like a display reviewer measuring contrast ratio, versus a manufacturer's claimed contrast ratio.

How exactly would you measure it? Would you use the pursuit camera technique at different speeds (pixels per second), and find the threshold at which the captured image showed a slight trail, and then derive the MPRT?
spacediver is online now  
post #15 of 72 Old 09-17-2013, 10:49 AM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

How exactly would you measure it?
There are so many ways to measure MPRT, much like measuring contrast ratios.
-- more subjectively -- For basic more-subjectively-measured MPRT measurements, you don't necessarily need a pursuit camera. You can use a resolution test pattern, like a moving test pattern (but won't be as accurate an average). A caveat needs to be added that when the test pattern looks perfect, the measured value is only a worst-case value. For example, a display with "1080 lines of motion resolution" (if the test pattern caps out at 1080 lines) may actually be MPRT 0.1ms, 1ms, or 2ms....
-- more objectively -- For more objectively-measured MPRT measurements meeting a specific standard (e.g. average MPRT, based on common picture material), a standardized test pattern along with a pursuit camera is recommendable.

For pursuit-camera-free MPRT tests, there are several ways to go about it.

One highly experimental example (not yet refined, not easy to use yet) is the TestUFO MPRT test pattern, at www.testufo.com/#test=mprt .... Try this on a common computer LCD, like the one you're currently using. (Make sure the animation runs at your refresh rate, 60 frames per second, or it won't work) .... Chrome works best. Safari usually works, as does IE10 and later. So does Opera 15 and later. You speed up/slowdown until the white squares are the same size as the black squares by adjusting the speed up and down. (If the test pattern is too slow the white squares are too small, and if the test pattern is too fast the black squares are too small. You adjust until there's an equilibrium. )

There are several limitations with this beta experimental pattern -- it only tests white/black MPRT by default. To get proper MPRT, you need to test several different color transitions and derive an average. Also, this test pattern works poorly on displays that use very unclean transitions (e.g. temporal dithering effects) and poorly on displays that use interpolation. For easier testing, test this test pattern on native-refresh-rate LCD's. It also reportedly works very well on DLP projectors too. Also, the test pattern is hard to use.

However -- the point is that it shows, today that various test patterns CAN be invented, that does manage to accurately measure MPRT (within a 10%-20% error margin), using the human eye alone. There is some subjective error injected into it, and it becomes more inaccurate the faster the MPRT is...

You can also use traditional moving resolution patterns, to a certain limit. The "lines of motion resolution" doesn't define a specific motion speed. But once the motion speed is known and the mapping is known, you can convert the numbers to "milliseconds of motion resolution" if certain variables are known, such as knowing that ther are "300 lines of motion resolution" for a specific test pattern of a specific motion speed, for a specific 60Hz sample-and-hold display. That can be assigned an MPRT of 1/60 = ~16.7 (A scientifically-perfect 60Hz sample-and-hold display has an MPRT of exactly 1/60.) Then thereafter for that specific test pattern at the same speed, the numbers can be proportionally mapped to MPRT's once a specific "lines of motion resolution" maps to a specific MPRT. (Or whatever preferred name is -- "persistence" or the inverse value, "motion equivalent ratio", etc)

The caveat is that once the test pattern caps out, you're not able to measure lower MPRT's by human eye alone, except through pursuit camera photography or scientific equipment. I have done tests that shows that test patterns that are capable of meaturing MPRT's down to approximately 1ms by human eye -- are actually possible/feasible, but requires fast-moving single-pixel-width lines in custom test patterns. Beyond this, begins to definitively require pursuit camera and scientific equipment...

I feel, there's going to be a long period of test pattern innovation for motion resolution tests. (ability to measure ultra high motion resolution, while still being user friendly, etc)

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #16 of 72 Old 09-17-2013, 12:15 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
ideally, a photodiode-based instrument would do the trick, but that's not at all convenient smile.gif

A camera pursuit system coupled with automated image analysis would work in principle. For example, the test object could be a white square moving rapidly across a black background. Once a photo is taken, it would be a simple matter to code an algorithm that measures the amount of motion blur on the photo (and the code could also verify the alignment of the pursuit track). The problem is that there is variation among cameras and their calibration (shutter speed, etc.) that might introduce noise. Plus it isn't that convenient for most end users, although major review sites could be given training.

A behavioural measure may be more promising, although this would have to be carefully thought out. I'm thinking along the lines of a identification/discrimination task whose difficulty can be precisely titrated (and which also becomes difficult/impossible with a certain amount of motion blur). You can then find the difficulty at which, say, an observer achieves 75 percent correct threshold.

For example, You could have a rotating landolt C, where you can adjust the gap width by a specified amount. The task would be to determine whether the target has a gap or doesn't have a gap (and on half the trials, the target would be a rotating circle (which has a gap width of 0). There will be two things that modulate the task difficulty: the gap width, and the rotational speed relative to the MPRT. It would be a simple matter to program this on matlab psychtoolbox (or the free equivalent of octave), where you have direct access to the frame buffer. It would also be a simple matter to program the experiment so that a 75 % threshold is obtained.

The crucial challenge in the behavioural approach is to design the task such that variance in performance is due solely to MPRT issues rather than individual differences in visual acuity and motion perception.

edit: Upon reflection, a better way may be to just start with a large gap, and instruct the observer to follow the gap as it goes around, and press a button to reduce the gap width incrementally until it disappears.
spacediver is online now  
post #17 of 72 Old 09-17-2013, 01:45 PM
AVS Special Member
 
Doug Blackburn's Avatar
 
Join Date: May 2008
Location: San Francisco - East Bay area
Posts: 3,453
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 19 Post(s)
Liked: 226
I still don't get it. How are you going to establish ONE number representing blur performance when the blur is different for different colors and luminances? It's just not possible.

Hypothetical example with made up numbers using a 0-100 scale with 0 being no blur and 100 being "a ton" of blur however you want to define a ton... say 25% of the width of the TV

Blur of 30% white on 10% white is 80
Blur of 70% white on 70% blue is 30
Blur of 100% white on 0% white is 40
Blur of 50% magenta on 50% white background is 60
Blur of 90% magenta on 90% white is 10
Blur of 100% red on 100% cyan is 10
Blur of 30% red on 30% cyan is 50
Blur of 60% green against 20% white is 90
Blur of 100% green against 100% blue is 20


And those numbers were derived with the display set for 35 fL for 100% white. If you change the display to 50 fL for 100% white, all the blur numbers change. If you change the display to 30 fL all the blur numbers change again. And what happens when you measure vertical blur, diagonal blur, and horizontal blur and the numbers are quite different for each direction of motion? (so if the blur pattern moves in a circular path, you'd measure a different amount of blur every at each point in the circle).

And the next variable is speed... slow motion tends to have less blur than fast motion so how do you account for the speed of motion trying to establish 1 number to describe blur performance? The amount of visible blur is DRAMATICALLY higher at high speeds vs, low speeds, for example.

No one number tells you ANYTHING useful about how the display works re. blur. Because the blur is different under different conditions. And no one "blur test" can describe the amount of blur on any given TV. So try one more time to convince me that there is a single number that will describe this hypothetical (but fairly realistic) blur performance.

Since blur is dependent on large numbers of variables, it seems impossible to evaluate blur without evaluating all the variables and that just plain takes a lot of time and a lot of different patterns.

You made the point that there's an order of magnitude difference... yes, POSSIBLY. But not when you are comparing two, say 120 Hz, LCD panels from the same model year... then you are going to have tiny differences/variations in blur performance. You will only see an order of magnitude difference when you compare some disparate technologies like DLP vs 120 Hz LCD. Or a 60 Hz CCFL LCD from 2005 to a 240 Hz strobing LED LCD from 2013. There is no argument that you would have an order of magnitude difference in some conditions. But nobody cares much about comparing a 2005 LCD to a 2013 top of the line LCD. People will want to know what the best 2013 LCDs are (in regards to low blur) and that's where the 2% or 5% differences are going to appear. But the differences could be dramatic for some combinations of luminance and color and insignificantly different for other combinations. So you just CANNOT assign one number to blur performance. Trying to characterize the overall blur performance of a TV in terms of 0-100% white, 0-100% red, 0-100% blue, and 0-100% green and all combinations of luminance gray and color is going to take a TON of time because there is NOT one pattern that will provide the (right) answer for the TVs overall blur performance. Even if you limit comparison points to 0, 25%, 50%, 75% and 100%, that's still a whole huge PILE of evaluation when you consider gray, red, green, blue, cyan, magenta, and yellow at those 5 levels in all possible combinations.

"Movies is magic..." Van Dyke Parks
THX Certified Professional Video Calibration
ISF -- HAA -- www.dBtheatrical.com
Widescreen Review -- Home Theater & Sound
Doug Blackburn is offline  
post #18 of 72 Old 09-17-2013, 02:28 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Doug Blackburn View Post

I still don't get it. How are you going to establish ONE number representing blur performance when the blur is different for different colors and luminances? It's just not possible.

As Mark indicated, a single value averaged across different experimental conditions may effectively solve this problem. Preliminary testing could yield some intelligent color contrasts and luminances to use as the experimental conditions. There's no reason a value of spread (plus or minus one standard deviation, for example) couldn't be included. In fact, a consumer who cares deeply about motion resolution would likely want access to both those values.
Quote:
Originally Posted by Doug Blackburn View Post

And what happens when you measure vertical blur, diagonal blur, and horizontal blur and the numbers are quite different for each direction of motion? (so if the blur pattern moves in a circular path, you'd measure a different amount of blur every at each point in the circle).

Again, different tests could be devised to account for this (and pilot studies could help determine whether blur varies significantly as a function of direction of motion)
Quote:
Originally Posted by Doug Blackburn View Post

And the next variable is speed... slow motion tends to have less blur than fast motion so how do you account for the speed of motion trying to establish 1 number to describe blur performance? The amount of visible blur is DRAMATICALLY higher at high speeds vs, low speeds, for example.

The important thing is the MPRT. MPRT doesn't change as a function of speed, but blur will. Remember, we are not interested in reporting blur, but MPRT. Blur will change as a function of speed, but so long as we know the speed, and can measure the blur, we can derive MPRT, and that will remain constant across different speeds (someone correct me if I'm wrong here).

Quote:
Originally Posted by Doug Blackburn View Post


You made the point that there's an order of magnitude difference... yes, POSSIBLY. But not when you are comparing two, say 120 Hz, LCD panels from the same model year... then you are going to have tiny differences/variations in blur performance. You will only see an order of magnitude difference when you compare some disparate technologies like DLP vs 120 Hz LCD. Or a 60 Hz CCFL LCD from 2005 to a 240 Hz strobing LED LCD from 2013. There is no argument that you would have an order of magnitude difference in some conditions. But nobody cares much about comparing a 2005 LCD to a 2013 top of the line LCD. People will want to know what the best 2013 LCDs are (in regards to low blur) and that's where the 2% or 5% differences are going to appear. But the differences could be dramatic for some combinations of luminance and color and insignificantly different for other combinations. So you just CANNOT assign one number to blur performance. Trying to characterize the overall blur performance of a TV in terms of 0-100% white, 0-100% red, 0-100% blue, and 0-100% green and all combinations of luminance gray and color is going to take a TON of time because there is NOT one pattern that will provide the (right) answer for the TVs overall blur performance. Even if you limit comparison points to 0, 25%, 50%, 75% and 100%, that's still a whole huge PILE of evaluation when you consider gray, red, green, blue, cyan, magenta, and yellow at those 5 levels in all possible combinations.


It would certainly be interesting to do some pilot work to test the variation at these different conditions and among different displays. You certainly want to avoid interpolation errors, so you don't want to skimp on the number of conditions too much. But it could well turn out that with well designed tests, it could take under 15 min to get a thorough evaluation (possibly WELL under 15 min).

I do agree that if there is significant variation, you want to report, at the very least, a measure of spread.
spacediver is online now  
post #19 of 72 Old 09-17-2013, 02:36 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Doug, I agreed with most of what you said.
But you missed the point: I am not disagreeing on how it's impossible to have one X value of say (contrast ratio, blur value).

The key detail is that motion resolution measurement is no more or less legitimate than contrast ratios.
Quote:
Originally Posted by Doug Blackburn View Post

I still don't get it. How are you going to establish ONE number representing blur performance when the blur is different for different colors and luminances? It's just not possible.
Likewise, for contrast ratios. How are you going to establish ONE number representing contrast ratios, when the contrast ratio is different for different variables (e.g. ambient light, environment, ANSI, black/white, checkerboard of different sizes, etc)? It's just not possible.
Quote:
Originally Posted by Doug Blackburn View Post

Hypothetical example with made up numbers using a 0-100 scale with 0 being no blur and 100 being "a ton" of blur however you want to define a ton... say 25% of the width of the TV

Blur of 30% white on 10% white is 80
Blur of 70% white on 70% blue is 30
Blur of 100% white on 0% white is 40
Blur of 50% magenta on 50% white background is 60
Blur of 90% magenta on 90% white is 10
Blur of 100% red on 100% cyan is 10
Blur of 30% red on 30% cyan is 50
Blur of 60% green against 20% white is 90
Blur of 100% green against 100% blue is 20
No disagreement. Again, no less illegitimate than contrast ratio. Also, new test patterns can slowly level (not perfectly) the playing field. Hypothetically:

Measured contrast ratio during bright environment is 200:1
Measured contrast ratio during dim environment is 2000:1
Measured contrast ratio during black/white splitscreen is 1500:1
Measured contrast ratio, total dark room, during ANSI standard checkerboard is 1000:1
Measured contrast ratio, total dark room, during fine checkeboard (8x8 matrix) is 500:1
Measured contrast ratio, total dark room, during ON/OFF is 50,000:1
Measured contrast ratio, average family room, during ANSI standard checkerboard is 500:1
Measured contrast ratio, average family room, during fine checkeboard (8x8 matrix) is 250:1
Measured contrast ratio, average family room, during ON/OFF is 10,000:1
Measured contrast ratio, dark black painted room, ANSI standard checkerboard is 1200:1
Measured contrast ratio, dark white painted room, ANSI standard checkerboard is 800:1
etc.

Again, my point still stands: Motion blur values are no more/less legitimate than contrast ratios.
However, the way blur measurements are not as good as they could become today.

From this perspective, even you would probably agree that "lines of motion resolution" is less legitimate than "milliseconds of motion resolution". Both have their problems and variables, but the former has a _superset_ of problems over the latter.
Quote:
Originally Posted by Doug Blackburn View Post

And those numbers were derived with the display set for 35 fL for 100% white. If you change the display to 50 fL for 100% white, all the blur numbers change. If you change the display to 30 fL all the blur numbers change again. And what happens when you measure vertical blur, diagonal blur, and horizontal blur and the numbers are quite different for each direction of motion? (so if the blur pattern moves in a circular path, you'd measure a different amount of blur every at each point in the circle).
All legitimate factors that can improve future test patterns.
Does not make MPRT any less legitimate than the variables that interfere with contrast ratios.
Based on this information, I even think you and I unamiously agree that today's "lines of motion resolution" is _even_ less suitable of a motion resolution.
Quote:
Originally Posted by Doug Blackburn View Post

And the next variable is speed... slow motion tends to have less blur than fast motion so how do you account for the speed of motion trying to establish 1 number to describe blur performance? The amount of visible blur is DRAMATICALLY higher at high speeds vs, low speeds, for example.
MPRT is speed independent.

On a scientifically perfect display, 1ms equals 1 pixel of motion blur for every 1000 pixels/sec
That means for 4ms MPRT
2 pixels of motion blur during 500 pixels/sec
4 pixels of motion blur during 1000 pixels/sec
8 pixels of motion blur during 2000 pixels/sec

It can vary when a display is doing some strange tricks (e.g. less-precise interpolation for faster speed motion), but when the motion is "pure", the MPRT is always a constant for clean frames on a clean-transition/clean-strobed displays.

On certain displays, the blur is a constant in all motion vectors
(we're excluding variables caused by interpolation here, since we don't use interpolation for computers / gaming):
-- OLED strobing
-- full-strobe backlight LCD's (not zone-based / scanning backlight LCD's)
-- black-frame insertion on DLP's
-- CRT displays

Recent motion surveys have shown a surprising co-relation between MPRT and what is seen by human eye on several of the newer display technologies that display clean persistence. Some display technologies now exist today, that manage to reach >90% of a scientific perfect match to MPRT, when running the ideal situation of framerate=Hz motion.
Again, I don't even consider blur contributed by interpolation, or blur contributed by compression, as we don't even use that for computers graphics and for gaming, so the blur variables become a little bit purer here in this respect, as well.

Just like measured ANSI contrast ratio is more reliable to what is seen, as an inky-black-screen technology, than for front projection (due to room diffusion).
Quote:
Originally Posted by Doug Blackburn View Post

No one number tells you ANYTHING useful about how the display works re. blur.
Agreed.
Same problem for contrast.
But, "lines of motion resolution" is a worse problem than "milliseconds of motion resolution"
Quote:
Originally Posted by Doug Blackburn View Post

Because the blur is different under different conditions. And no one "blur test" can describe the amount of blur on any given TV. So try one more time to convince me that there is a single number that will describe this hypothetical (but fairly realistic) blur performance.
Agreed.
Same problem for contrast.
But, "lines of motion resolution" is a worse problem than "milliseconds of motion resolution"
Quote:
Originally Posted by Doug Blackburn View Post

Since blur is dependent on large numbers of variables, it seems impossible to evaluate blur without evaluating all the variables and that just plain takes a lot of time and a lot of different patterns.
Agreed. But that's not my point.
Quote:
Originally Posted by Doug Blackburn View Post

You made the point that there's an order of magnitude difference... yes, POSSIBLY. But not when you are comparing two, say 120 Hz, LCD panels from the same model year...
The chart containing order of magnitude difference is on the same 120Hz LCD, configured to very different settings (different refresh rates and different backlight settings).

The best 120Hz monitor has an order of magnitude difference to the worse same-year 120Hz monitor. Several 120Hz monitors, such as the new 2013 AOC 120Hz monitor has no strobe backlight, and thus are bottlenecked by the lack of a strobe backlight.
Quote:
Originally Posted by Doug Blackburn View Post

Then you are going to have tiny differences/variations in blur performance. You will only see an order of magnitude difference when you compare some disparate technologies like DLP vs 120 Hz LCD.
Not true.
Again, we're talking about computer graphics and game material, not videos and movies.
Different people have different sensitivities.
Quote:
Originally Posted by Doug Blackburn View Post

Or a 60 Hz CCFL LCD from 2005 to a 240 Hz strobing LED LCD from 2013.
Agreed.
Quote:
Originally Posted by Doug Blackburn View Post

There is no argument that you would have an order of magnitude difference in some conditions. But nobody cares much about comparing a 2005 LCD to a 2013 top of the line LCD.
Re-read.
There were certain same-year 2013 LCD's have order of magnitude differences relative to each other, and in various different settings.
Yes. Surprised? Come fly here, I'll demonstrate. I'll set aside the time. Riverdale area in Toronto, Ontario.
Quote:
Originally Posted by Doug Blackburn View Post

People will want to know what the best 2013 LCDs are (in regards to low blur) and that's where the 2% or 5% differences are going to appear.
False. The differences are much larger.
We're not talking about video and movie material where display motion blur is easily hidden by source/compression based motion blur.
Again, I give you a standing invitation for a demonstration of the massive blur differences on a modern strobe-backlight LCD display in various different modes of operations, from the perspective of computer usage.

For example, see PHOTOS: 60Hz vs 120Hz vs LightBoost. Same screen. Same model year display.
There are also testimonials from the large differences that the computer monitor world recently witnessed.

Also, we're limiting ourselves to interpolation-free motion-blur-reduction technologies.
Interpolation adds input lag, making it unsuitable for Game Mode.
Until recently, it was not possible to have high-efficiency low-latency interpolation-free
Quote:
Originally Posted by Doug Blackburn View Post

But the differences could be dramatic for some combinations of luminance and color and insignificantly different for other combinations. So you just CANNOT assign one number to blur performance.
Agreed.
Same problem for contrast.
Quote:
Originally Posted by Doug Blackburn View Post

Trying to characterize the overall blur performance of a TV in terms of 0-100% white, 0-100% red, 0-100% blue, and 0-100% green and all combinations of luminance gray and color is going to take a TON of time because there is NOT one pattern that will provide the (right) answer for the TVs overall blur performance. Even if you limit comparison points to 0, 25%, 50%, 75% and 100%, that's still a whole huge PILE of evaluation when you consider gray, red, green, blue, cyan, magenta, and yellow at those 5 levels in all possible combinations.
Agreed.
Same problem for contrast.

Again, we're not talking about measuring motion blur for video, for interpolation, or for movies.
We're talking about measuring motion blur for users who use their big screen with a computer or game system, in low-latency interpolation-free Game Mode.

Until recently, low-lag Game Mode blur-reduction was not feasible (e.g. enabling Motionflow in Game Mode)

Now we are hit during year 2013 of a special situation where SAME YEAR LCD televisions and computer monitors can have several-times differences (and up to around an order-of-magnitude) differences in GAME MODE motion blur. For example, just compare the GAME MODE of a 2013 Samsung LCD (it doesn't *even* let you adjust any blur- reducing settings during game mode!), and the GAME MODE of a 2013 Sony LCD that has "Motionflow Impulse" enabled. Shocking motion clarity difference. Some situations measured half an order of magnitude. Ignore all blur-reducing modes that has more than 2-3 frames of input lag -- that's the stuff that uses interpolation (making interactive use unsuitable, such as gaming or computing)

This is approximately the panning clarity difference you see during ~60Hz versus ~240Hz (but with 240Hz simulated by backlight pulse width, one 1/240sec flash per 60Hz frame). Such major clarity differences in same-year HDTV's during Game Mode, is currently unprecedented territory, and many reviewers do not realize the importance of testing Game Mode motion blur, using full-resolution / full-progressive / full-framerate test pattern generators running at faster motion rates (e.g. 16ppf or 960 pixels/second).

Formerly, strobe/scanning backlights used to be force-bundled with interpolation (Existing scanning backlight technology that can't be disabled without also disabling interpolation). Interpolation is incompatible with computers and games, due to input lag -- taking 80-100 milliseconds from clicking a button and seeing something happen on the screen. We have an unusual situation where same-year LCD's have massive blur differences (DURING GAME MODE) to others.

Even regardless of all the above, nearly all the current blu ray motion tests don't push the limits of what can be detected by the human eye during computer graphics material and game material. (at least when pointed out -- e.g. many people don't see rainbows and many people don't see 3:2 pulldown, unless they are "trained" to notice them better). Blur Busters Blog get tens of thousands of visits per week from blur-sensitive people, so there's a quickly emergent market demand for better 21st century blur-measurement tools, at least in the computer/gaming audiences. Mainstream may not care, but they could certainly also appreciate comparision shopping via a blur measurement standard superior to "lines of motion resolution". Again, it's no less illegitimate than measuring contrast ratios. A widely agreed industry standard for measuring motion resolution (similar to "ANSI checkerboard") could eventually do the industry a good service over the long term, even if it is imperfect.

As of 2013 some huge computer-compatible LCD blur standouts are:
-- Sony's Motionflow Impulse on certain Sony TV's (dramatic horserun ahead of all other LCD's). People have reported being surprised by the clarity of 60fps gaming on these.
-- nVidia LightBoost (starting as little as $250) with an unofficial tweak and rave testimonials
-- Certain speciality LCD monitors
.........Eizo FDF2405W (MSRP $6999) -- strobe backlight documented in page 15 of PDF manual
.........Viewpixx Scientific LCD (MSRP >$20,000 from vpixx.com) -- ultra-low-diffusion high efficiency scanning backlight (7x improvement in motion resolution, just simply by turning it on -- with NO interpolation -- and low latency compatible with computer use)

All these LCD are sudden dramatic same-year computer-compatible motion-blur standouts (the worst display of the above has more than 3x the motion resolution of the best non-strobe-backlight LCD's still being put on the market in 2013 model year). This blur innovation happened so fast, so sudden, reviewers haven't even discovered yet -- except a few people like Blur Busters and certain select media coverage, especially since most of mainstream media doesn't even understand the principle behind strobe backlights (which are superior to scanning backlights). Also, CRT's are an excellent example of 1990's displays that have far better motion resolution than most of this decade's displays. We are witnessing gamers/computer users pleasantly surprised by the CRT-clarity upgrade in certain LCD's (see above list), in same-model year (beginning with some 2012 models).

Right now most people don't think it's important (right now).
Some have disadvantages (e.g. excessive flicker in certain 60Hz strobe backlights that outweigh its benefit during 60fps@60Hz HTPC gaming.)

But it's going to become increasingly important as time passes.
Especially as 120Hz gets more standardized in the coming decade (e.g. NHK 8K 120Hz) and makes pratical eye-comfortable interpolation-free strobing. Something to monitor.
Some people are far more motion-blur-sensitive than other people -- especially people who say they have have never seen a flat panel as clear as a CRT.

Even plasma/DLP/etc needs better motion tests. The famous NeoPDP panels have ultra-short-persistence phosphors, and actually produces far more motion resolution than a Kuro, especially noticeable in computer/gaming use cases, one Blur Busters user purchased NeoPDP's precisely because of motion resolution interests. And a recent discovery by some gamers, and confirmed by Blur Busters, is that you can quadruple the motion resolution of certain DLP projectors such as the low-end Optoma GT720, via a 120Hz+BFI trick. (You simply input a 120Hz native computer signal, while in 3D mode, even though you're not displaying 3D. In high-speed camera tests, the GT720 DLP projector adds undocumented black frame insertion during 3D mode to reduce crosstalk, but accidentally doubles motion resolution even further, above-and-beyond the doubling that just going to 120Hz does).

Anyway, today's "lines of motion resolution" patterns need quite a lot of improvement, and is not good enough in the modern era of computers and games connected to displays. They fall fall short in many areas.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #20 of 72 Old 09-20-2013, 10:36 AM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Just a clarification, "Motion Picture Respose Time" is actually called "Moving Picture Response Time" in the original science papers.
Some papers use the other term, and others use the other. However, they mean the same thing. Here's an excellent search on Google:

Google: Moving Picture Response Time
  1. It's resolution independent.
  2. It's motion speed independent. (most displays have the same MPRT at all motion speeds and motion vectors)
  3. It's more test pattern independent.
  4. It's future proof. 1080p, 4K, 8K, VR, etc.
  5. Improvement is unbounded. It doesn't cap out at a specific value (e.g. "1200 lines of motion resolution")
  6. It easily covers the faster motion speeds often seen in video game use, an increasing use case of displays.

Easy consumer motion test patterns needs to be developed over the coming years, to discontinue the "lines of motion resolution" lingo. Even disregarding the subjectiveness problem (like Contrast), "lines of motion resolution" just simply has a far bigger superset of problems far bigger than "milliseconds of motion resolution" or "MPRT" or "moving picture response time" (which is NOT the same measurement standard as "LCD response time"). It may take time (e.g. Blu Ray becoming obsoelte), but it's time to begin migrating to a a future proof motion resolution standard.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #21 of 72 Old 09-20-2013, 10:41 AM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post


Easy consumer motion test patterns needs to be developed over the coming years, to discontinue the "lines of motion resolution" lingo.

On this note, what are your thoughts on the rotating landolt C idea? (see my edit in bold earlier)
spacediver is online now  
post #22 of 72 Old 09-20-2013, 10:50 AM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

On this note, what are your thoughts on the rotating landolt C idea? (see my edit in bold earlier)
It wouldn't work very well:

1. Non-constant-vector motion do not test limits of human vision acuity.
Panning is the most common use case during gaming (panning, turning, strafing, etc). Human eyes are more accurate at tracking during linear motion. Given the motion test pattern speeds for accurately testing videogame motion blur (~16ppf or 960 pixels/second), tracking 960 pixels/second in rotating motion is much harder than tracking 960 pixels/second linearly. Imperfections in rotating motion will be harder to see than imperfections in linear motion, so I'm against nonlinear motion resolution tests, at least for motion tests that are intended to be seen by the human eye. Better to test linear motion tests in multiple axes and average the results (to accomodate displays that have differences in vertical motion blur versus horizontal motion blur -- very rare -- but could happen). Without seeing this specific test first, it doesn't sound like it can test the limits of human vision acuity for common linear motion.

2. It is definitely not a motion resolution test.
Aliasing effects of a rotating gap can still often be detected even at less than 1 pixel thickness. The human detection of this would not be representative of display MPRT. So the Rotating Landolt C test would not be a motion resolution test, but a "can you detect the faint rotating pixelly artifact?" -- given sufficient high contrast and certain rotating speeds. One can easily detect a 0.1 pixel gap in a static pattern if it's antialiased -- it's simply a faded one-pixel thick artifact. Blended/antialiased 0.1 pixels -- e.g. a 0.1 pixel thick white line drawn on a black background, would show up as a dim grey 1 pixel thick line. So it is not a motion resolution test anymore, but "ability to see dim pixels" test. (e.g. With blending from antialiasing for a 256-greyscale color space (8-bit) you would see grey "#26 out of 256" for 0.1-pixel-thick horizontal white line, "grey #3 out of 256" for a 0.01-pixel-thick horizontal white line on back background) The rotating Landolt C would just be a temporal (moving) equivalent of this test -- basically "ability to see fast-moving dim pixels" test, and definitely not a "motion resolution test".

For simple one-pass tests meant to be interpreted by human eye, horizontal motion should still continue to be standard
For this, I'm temporarily setting aside the subject of fully scientific multi-axis motion tests (one could view this as a rough analog to detailed scientific measurement of contrast ratios based on hundreds of display measurements at all points/corners of the display, because contrast ratio varies due to display nonunformities).

The point is keeping it simple for the human vision. The huge majority of displays have already proven to have identical motion resolution in all axes of motion during computer graphics (LCD's, CRT's, DLP's, plasmas), if interlacing/deinterlacing becomes excluded from the equation. For modern computer and gaming use cases, we're not using deinterlacing and we're not using interpolation. So let's simplify on the most common use case: Panning along the long axis of the display. It mimics the common motion cases. It uses the axis where there's more time for humans to detect motion blur. For important science, you can come up with omnidirectional test patterns or average multiple vectors (e.g. 360 tests, at every degree, and average/graph the result). But simple horizontal motion should suffice since it's already confirmed that MPRT is usually constant or near-constant during all motion vectors (at least when unaffected by interlacing/compression/interpolation/deinterlacing/etc -- interlacing/deinterlacing is the chief cause of MPRT divergences based on motion axis -- and computer/games don't have this problem)

So any "simplified" human-eye-based Motion Picture Response Time (MPRT) tests invented in the future, would probably be horizontal motion tests of various kinds and inventions. As easy as today's tests, but instead measuring MPRT instead of measuring "lines of motion resolution". Measuring of MPRT by human eye is best done by interactive motion tests (adjust speed until a motion blur equilibrium to a reference comparison image is reached, for example -- still somewhat subjective but no less so than current motion tests and no less so than Contrast measurements). I have successfully visualized experimental tests that could theoretically statically be recorded onto a Blu-Ray. Care must be taken to make sure compression artifacts don't interfere, and the process of deinterlacing 1080i->1080p doesn't interfere (alas, it unfortunately will, especially at motionspeeds representative of computer/gaming motion). Blu-Ray motion tests, will have extreme difficulty being able to test motion resolution for computer use cases / gaming use cases, unless 1080p/60 is standardized into Blu-Ray (for 1:1 pixel mapping, unadulterated by interlacing / deinterlacing interference to motion resolution).

Multipass tests and scientific motion tests can still continue to use all motion axes
This doesn't mean other tests can be invented, especially for measuring equipment (not just human eye) for more accurate/scientific measurement of MPRT. Just simply all axes of motion
Likewise, some display researchers / scientists often do multipass contrast readings. e.g. Instead of measuring only the screen center or one square, you measure hundreds of points in the display (uniformity test) and get an average contrast ratio, min / max, etc.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #23 of 72 Old 09-20-2013, 11:20 AM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon 
Imperfections in rotating motion will be harder to see than imperfections in linear motion, so I'm against nonlinear motion resolution tests, at least for motion tests that are intended to be seen by the human eye.

Makes sense.
Quote:
Originally Posted by Mark Rejhon 
Aliasing effects of a rotating gap can still often be detected even at less than 1 pixel thickness. The human detection of this would not be representative of display MPRT. So the Rotating Landolt C test would not be a motion resolution test, but a "can you detect the faint rotating pixelly artifact?" -- given sufficient high contrast and certain rotating speeds, I can easily detect a 0.1 pixel gap. Aliased 0.1 pixels -- e.g. a 0.1 pixel thick white line drawn on a black background, would show up as a dim grey 1 pixel thick line. So it is not a motion resolution test anymore, but "ability to see dim pixels" test. The rotating Landolt C would just be a temporal (moving) equivalent of this test -- basically "ability to see fast-moving dim pixels" test, and definitely not a "motion resolution test".

A linear equivalent may work. So you have a white rectangle split in half by a vertical gap. The white rectangle moves across the screen and user adjusts gap size until the rectangle takes on a uniform appearance.

I think you can achieve sub pixel accuracy with a gaussian filter on the inner edges of the rectangle. So at the gap, it wouldn't suddenly jump from white to background color and back to white, but would rather change luminance gradually. You can shift the mean of this filter across space with arbitrary precision. This is a controlled aliasing effect if you will.

Now when the gap width decreases, the gap itself gradually approaches the luminance of the rest of the rectangle. With zero motion blur, you'd have to have the means of the filter (which themselves are centered on the theoretical inner edges of the rectangle) occupy the same spatial position. With a bit of motion blur, the blur itself would do a bit of filling in.
spacediver is online now  
post #24 of 72 Old 10-04-2013, 12:31 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
EDIT:
Quote:
-- Sony's Motionflow Impulse on certain Sony TV's (dramatic horserun ahead of all other LCD's). People have reported being surprised by the clarity of 60fps gaming on these.
Whoops, that was a bad link. I've fixed the link to point to www.blurbusters.com/sony-motionflow-impulse

REPLY:
Quote:
Originally Posted by spacediver View Post

A linear equivalent may work. So you have a white rectangle split in half by a vertical gap. The white rectangle moves across the screen and user adjusts gap size until the rectangle takes on a uniform appearance.
Quote:
Now when the gap width decreases, the gap itself gradually approaches the luminance of the rest of the rectangle. With zero motion blur, you'd have to have the means of the filter (which themselves are centered on the theoretical inner edges of the rectangle) occupy the same spatial position. With a bit of motion blur, the blur itself would do a bit of filling in.
I already do an indirect variant. It's roughly a method of www.testufo.com/eyetracking or www.testufo.com/mprt which is indirectly based off www.testufo.com/chase where the blur trail touches the pixels behind it. Shorter distance vs longer distances between chasing pixels. Faster motion means more motion blurring. Etc. I'm already familiar with this.

I have an idea of how to simplify this; but it can be quite challenging due to strobe backlight (sharp falloff of light) versus phosphor decay (slow falloff of light) which can muddy subjective observations to an extent. What is needed, is a standardization (e.g. 50% cutoff, e.g. a blending trigger point), much like how contrast is standardized from ANSI versus full on-off. Even though panels may have different contrast for different parts of the display than center of display, etc. The art of coming up with a motion blur equivalence number, can be much, much superior than it was in the past.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #25 of 72 Old 10-04-2013, 03:25 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
very cool demo.

By standardization, I take it you mean a reliable way of using the human visual system (rather than an image from a pursuit camera) to measure a standard amount of motion blur, right?

In the case of your checkerboard pattern, this standardization would be a visual estimation of when the black and white rectangles are the same size, right?

In my idea, it would be when the luminance of the gap matches the luminance of the rest of the rectangle.

Would the shape of the decay function complicate things in both the above paradigms?
spacediver is online now  
post #26 of 72 Old 10-04-2013, 05:30 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

By standardization, I take it you mean a reliable way of using the human visual system (rather than an image from a pursuit camera) to measure a standard amount of motion blur, right?
Actually, I meant standardizing the cutoff points. For example. Do we cutoff phosphor decay's contribution to motion blur, when the phosphor has decayed 50%? 90%? 99%? We're talking about situations where the display is not producing a clean sample-and-hold effect (which allows for more perfect equivalence between what the human/camera saw, and the numbers that are recorded for a "Motion Equivalence Rating" number, or a "Motion Blur in milliseconds" number.

For example, if a strobe backlight takes 1ms to gradually turn on, 1ms to brightly illuminate, and 1ms to gradually turn off, that won't be creating 3 milliseconds of motion blur. Photographically, it looks like 2 milliseconds of motion blur (2 pixels for every 1000 pix/sec motion). Here, I think the standardization point, from the perspective of motion blur, is 50%. So from the 50% cutoff point of the 1ms rise, to the 50% cutoff point to the 1ms fall, is equal to the time period from 0.5ms through 2.5ms of the 3-millisecond long "gradual soft strobe". So that 3ms strobe is actually creating roughly 2ms of motion blur, and gets assigned a MPRT of 2.0ms, or a Motion Equivalence Ratio of 500. (1/MPRT). Obviously, this is an imperfect conversion, but no less imperfect than the art of contrast ratio claims/reality/etc.

Various things can fuzzy-up the sample-and-hold effect, either at the ramp-up period and ramp-down period. For example, plasma subfield ramping up, CRT phosphor decay, plasma phosphor decay, etc).

Also in regards to pursuit camera photography, the accuracy of the pursuit camera photographs, as imperfect/perfect they may be, they aren't the issue here. The decay effects are accurately captured too in pursuit cameras, as relative to seen by human eye. When averaged over a ~50ms timescale (e.g. 1/20sec or so) capturing a few refreshes to average temporal effects out, the motion blurring captured in a pursuit camera have an uncanny visual resemblance. You easily compare what you saw with your eyes at www.testufo.com/ghosting and the photos already published at Blur Busters (the LCD Motion Artifacts article, the 60Hz vs 120Hz vs LightBoost, or the newly published LCD Overdrive Artifacts article.) Everyone who compared, says the photographs have an uncanny ressemblance of what they saw with their human eye, if they're comparing from the same league of monitors that the pursuit camera photos were taken from. It's not a perfect capture as everyone's vision system is different, but they are far more accurate representations.
Quote:
In the case of your checkerboard pattern, this standardization would be a visual estimation of when the black and white rectangles are the same size, right?
Essentially yes.
(Or visual estimation from a pursuit camera photograph of the same test -- e.g. http://www.testufo.com/mprt#easteregg=1&pursuit=1 -- when done this way, the camera image looks exactly the same as what you saw by human eye).

Right now, the test is too complex, but I've successfully found ways to simplify this further. The test is only good up to displays as fast as ~4ms to ~8ms MPRT, displays that are faster than that, will require motion speeds too fast to easily follow, but I have done experimental test patterns that successfully allowed measuring 1ms vs 2ms MPRT by human eye.
Quote:
In my idea, it would be when the luminance of the gap matches the luminance of the rest of the rectangle.
It doesn't work well that way. Sample-and-hold motion blur can't be reliably measured that way, since the gap can be shorter than the length of a refresh cycle, and thus the sample-and-hold motion blur is much larger than that gap. It comes to the point where the gap simply fades and fades, but never completely disappears, until the motion is too fast, and the resulting number has a huge error.

Unlike the MPRT test, which has an error margin of only 10% on computer monitor LCD's, between bechmarking from human eye and scientific equipment measurements. You really want the motion blur to perfectly fill the gap, so you need to make the gap large enough to cover a full refresh cycle for a motion speed. For a 16 pixel gap, you end up having a 16 pixels per frame (960 pixels/sec) for a 60Hz sample-and-hold display since the pixel is being continuously shining as your eyes track over a linear 14 pixels over a timespan of 1/60sec.

Now if you've got a display that's using black frame insertion, e.g. 60Hz display with a 50%:50% dark cycle (like the Sony Trimaster OLED), then the 16 pixel gap fills itself at motion speeds of about 8 pixels per second! You adjust motion speed "Pixels Per Frame" until the gap fills entirely. 7 or 8 pixels of motion blurring instead of 16 pixels of motion blurring. If you try www.testufo.com/mprt on the Sony Trimaster OLED during it flicker mode, you'll see it gives you a MPRT of 7.5ms, and a Motion Equivalence Ratio slightly higher than 120, representing its already-measured 7.5ms of known persistence.

You immediately notice that the motion blur math is startlingly simple for clean-strobe displays (e.g. strobe LCD's, strobe OLED's)
Gap Size == (Duty Cycle) x Pixels Per Frame = Pixels Of Motion Blur == Near Perfect match
During 100% duty cycle, you use "1".
During 50% duty cycle, you use "0.5".
I say "Near perfect match" because the only error is caused by asymmetric pixel transition speeds (aka LCD ghosting/overdrive, which is nearly non-existent on modern LCD's). For 100% duty cycle, during 100% sample-and-hold, 100% brightness, non-PWM dimming.

For 50% flicker (black frame 50%:50% dark:bright), motion speed is halved to perfectly fill the gap with motion blur:
Gap Size == 1/2 Pixel Per Frame

For 25% flicker (black frame 25%:75% dark:bright), motion speed is halved to perfectly fill the gap with motion blur:
Gap Size == 1/4 Pixel Per Frame

See; the motion blur math is quite simple for illumination that resembles square wave (clean-strobe LCD's and clean-strobe OLED's. (near instantly to ON, continuous, then nearly instant OFF).

Let's limit eye tracking accuracy to only half a screen width per second, since not all humans can accurately track moving objects faster than that. So we need to shrink the gap to measure 1ms MPRT's. 1 pixel thick at 960 pixels per second gives the necessary accuracy to detect 1ms MPRT's.

A prototype MPRT test for LightBoost LCD's is http://www.testufo.com/#test=mprt-fast .... The fine lines becomes a sold block of grey at 8 pixels per frame during LightBoost 120Hz at LightBoost=100%. The benchmark is telling your MPRT 1.0ms (Try even multiples, as odd multiples start creating LCD inversion artifacts on some displays). The error margin is bigger, but less than 25% of the oscilloscope measurement (2.4ms strobe flash!). Now when you use LightBoost=10% (shorter strobe flash), the square becomes a solid block of grey at 12, 14 or 16 (depending on display). That's MPRT 1.0 to 1.4, very close to the actual oscilloscope measurements. (I recently discovered certain LightBoost LCD's actually use a shorter 1 milisecond strobe length during LightBoost=10%).

So, you see, I am using a very simple, scientifically provable math formula to display the MPRT numbers on these motion test patterns. Giving benchmark numbers to these is thusly, relatively simple. The chief complication is standardizing giving numbers to complex motion blur (see below)

One problem is inventing one motion test that can easily give benchmark numbers on all display technologies (plasma, DLP, LCD, CRT, etc). The decay effects creates a large complication.
Quote:
Would the shape of the decay function complicate things in both the above paradigms?
Definitely. Decay manifests itself as ghosting, streaking,
CRT: green ghosting
LCD: trailing ghosts or overdrive artifacts
Plasma: yellow ghosting, noisy/banded blur trail

(They're both seen, and they're capturable in an accurate manner using pursuit camera. See my new article about LCD Overdrive Artifacts. If you have an ASUS 120Hz monitor with LightBoost turned off, you can compare what you saw with your eyes at www.testufo.com/ghosting with the photographs, and see the closer-than-expected resemblance)

Again, the difference of the blur inside a pursuit camera photograh, and the blur of what the human eye saw, isn't the issue here -- we're using well-controlled motion to simulate the ideal motion scenarios (similiar to computer game motion on an ultrapowerful GPU during framerate-locked situations (ala the butter smooth platformer scroll effect of perfectly fluid game motion). In these situations, it is quite possible to see motion blur from even a 1ms MPRT difference (Important reminder: MPRT measurement is not GtG measurement, by the way).

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #27 of 72 Old 10-04-2013, 06:07 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Here's proof of how accurate pursuit camera photographs are.

1. View http://www.testufo.com/mprt#easteregg=1&pursuit=1 on a recent fast 2ms 60Hz LCD, 100% Brightness

2. Compare what you saw with your eyes with this actual pursuit camera photo:



See? The checkerboard is photographically captured successfully with a pursuit camera! The size of motion blur is virtually IDENTICAL between what you saw with your eye, with what the camera captured. (Camera was set to 1/15sec exposure speed, capturing 4 refreshes in a row, as the camera tracked the moving pattern)

(Note: Minor variations between different 60Hz LCD's, so you're actually comparing your LCD with a pursuit camera photo of my own LCD, but most LCD, set to 60Hz without interpolation and without strobing, nowadays look almost identical in terms of motion blur. The comparision becomes even more accurate if you own exactly the same LCD that I have)

For more information about the low-budget pursuit camera technique, see HOWTO: Pursuit Camera with Blur Busters Motion Tests.
______

Now back to spacediver's inquiry two posts above:

The issues I'm solving is unrelated to human vision blur versus pursuit camera blur (I consider that a 'reasonably' solved part of the equation right now) -- just view www.testufo.com/mprt on a plasma display and you'll understand better what I mean. On plasma, the squares are no longer square-shaped because of the plasma dithering & plasma yellow ghosting/deay. Yes, it is also accurately captured in pursuit camera photograph too! (plasma yellow blur and all). Even the plasma banding artifacts are captured in pursuit camera photographs (it's already actually done for scientific papers too). The only thing the photo doesn't capture is the flickeriness of the pixels (within-camera-shutter-speed termpral effects). Plasmas tend to roughly balance the luminance out approximately at MPRT=4ms ("Pixel Per Frame" motion speeds about 4x faster than a 60Hz LCD), but the squares aren't cleanly square shaped.

An essentially rephrased version of the question I am asking myself, is "how do we slightly more objectively assign motion blur equivalence numbers or MPRT values, to plasma motion measurements, accounting for these kinds of effects?" It becomes somewhat as subjective as the archaic "lines of motion resolution", but less subjectively so. I can see how standardizing this could become controversial (but less so than "lines of motion resolution" which, due to its limitations, can be more subjective than "milliseconds of motion resolution") The amount of motion blur found in colorful objects can be different than the motion blur of black-white, so the MPRT test needs to be repeated between common color combinations (e.g. red vs green, blue vs red, dim red vs yellow, etc) roughly evenly spread throughout the RGB spectrum, to come up with reasonably accurate MPRT / MER numbers that's far more scientifically accurate and objective than the current "lines of motion resolution" tests today.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
post #28 of 72 Old 10-04-2013, 07:29 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
I was thinking more along the lines of devising a test where, instead of deriving the MPRT/MER (which I still don't have a fluent grasp of!), you just find the threshold of that particular test.

So that way, if someone wants to compare two displays, they can do so by comparing the results from a specific test. Wouldn't that eliminate the problems of decay etc? I think perhaps I'm missing the whole point of your idea, but lemme see if I can grasp it by going back to your response here:
Quote:
It doesn't work well that way. Sample-and-hold motion blur can't be reliably measured that way, since the gap can be shorter than the length of a refresh cycle, and thus the sample-and-hold motion blur is much larger than that gap. It comes to the point where the gap simply fades and fades, but never completely disappears, until the motion is too fast, and the resulting number has a huge error.

Wouldn't the guassian filter deal with this problem? You can adjust the width of the gap by mathematically changing the location of the mean, and you can do so with arbitrary spatial precision. That way, you wouldn't need to speed up the motion to a crazy amount. You just keep the motion constant (at a reasonable speed that's trackable), and reduce the gap width until it turns completely white. You then have a number (gap width, or distance between means, with sub pixel resolution) that you can compare across displays.
spacediver is online now  
post #29 of 72 Old 10-04-2013, 08:21 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 785
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 66
Quote:
Originally Posted by Mark Rejhon View Post

Also in regards to pursuit camera photography, the accuracy of the pursuit camera photographs, as imperfect/perfect they may be, they aren't the issue here. The decay effects are accurately captured too in pursuit cameras, as relative to seen by human eye.

I wasn't questioning the accuracy of pursuit camera - just saying that it would be nice to have a test that didn't rely on one.
spacediver is online now  
post #30 of 72 Old 10-04-2013, 08:36 PM - Thread Starter
AVS Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 8,124
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 96
Quote:
Originally Posted by spacediver View Post

I was thinking more along the lines of devising a test where, instead of deriving the MPRT/MER (which I still don't have a fluent grasp of!), you just find the threshold of that particular test.
One simpler method to interpret MPRT ("Moving Picture Response Time") is that it is just simply a reinterpretation of "lines of motion resolution".
But in a way that can be measured more accurately, and can be recorded in a number that is test-pattern independent.
Firstly, you can even begin by just using today's "lines of motion reslution" test pattern for now.

-- If a display's known MPRT is 16.7ms, and that display known "lines of motion resolution" (for one specific Blu-Ray test pattern at one specific motion speed) is "300 lines of motion resolution", then it's easy to approximately translate between MPRT and "lines of motion resolution" that specific test pattern:
MPRT 16.67ms = 300 lines of motion resolution
MPRT 8.33ms = 600 lines of motion resolution
MPRT 4.16ms = 1200 lines of motion resolution
(If the test pattern caps out at 1200 lines of motion resolution, its ability is limited. Thus, for that specific pattern, you can't measure MPRT better than about 4ms, which happens to be conveniently near time of plasma phosphor decay of a very fast plasma phosphor.)

-- Now, if you had a different Blu-Ray pattern where it measures 200 on a fast 60Hz LCD in Game Mode (16.7ms MPRT), with different resolutions, caps out at 1080 lines, and runs at a different motion speed, you might have conversion numbers:
MPRT 16.67ms = 200 lines of motion resolution
MPRT 8.33ms = 400 lines of motion resolution
MPRT 4.16ms = 800 lines of motion resolution
MPRT ~3ms = 1080 lines of motion resolution
(caps out here)

Then you've got a 4K video file, with lines of motion resolution that goes all the way to 2000 (more dense lines), which creates this example:
MPRT 16.67ms = 500 lines of motion resolution
MPRT 8.33ms = 1000 lines of motion resolution
MPRT 4.16ms = 2000 lines of motion resolution
(cass out here)

Now, you begin to realize the problem; comparing different benchmarks of "lines of motion resolution" becomes problematic!
-- MPRT is test-pattern independent
-- MPRT is resolution independent
-- MPRT doesn't cap out. There's no upper or lower bound.

You can see then, here.
-- MPRT is more apples-to-apples comparision.
-- MPRT is future proof.
-- MPRT is more standardizable.

Also, we already scientifically know for MPRT:
-- 1ms of motion blur equals 1 pixel of motion blur for every 1000 pixels/second of motion
-- Example: That means 6ms of MPRT at 500 pixels/sec creates 3 pixels of motion blur
-- Example: That means 9ms of MPRT at 2000 pixels/sec creates 18 pixels of motion blur

Subjective versus Objective
-- MPRT can optionally be used with a pursuit camera.
-- Obviously, various inefficiencies such as temporal dithering/ghosting/rise/fall times, will come into play, but displays are getting better and better and getting so good that the motion blur measurements very accurately matches MPRT, especially on several display technologies.
-- LCD, OLED, DLP works great with the human-eye-based MPRT test, for example and the subjectively-observed numbers closely agree with objectively-measured numbers in these situations!
-- The situation of CRT and plasma becomes slightly complicated but apparently, when I run the numbers at "50% rise-fall" cutoff points, the numbers between objective-vs-subjective apparently come into closer harmony! The question is whether this is appropriate/good enough (aka non-controversial)

MPRT (full name "Moving Picture Response Time" (Google Search)) was originally invented for LCD's, but is not properly being used by monitor manufacturers because MPRT's actually give terrible numbers to LCD's once LCD's became far faster than 16ms. MPRT is *not* GtG transition measurement. Instead, monitor manufacturers are instead measuring the time it takes for a pixel to transition (GtG), rather than measuring the time of the sample-and-hold effect. Completely ignoring MPRT. However, it's apparently a good standard (The MPRT standard was mentioned to me -- by no less than an email from Raymond Soneira from DisplayMate) -- I was trying to find a standard way of measuring motion blur and that was one of them. I have grown to like the standard very much, and I have learned quite a lot in the last few months, with a more intimate understanding of motion blur. Although Raymond and I may have different approaches to motion testing, MPRT is quite a great standard that deserves a closer look. The MPRT standard is quite applicable to all displays, not just LCD's, and even retroactively applicable to CRT's. Yes, it's problematic, but no less problematic than the concept of "contrast ratios" (given our human visions' limited dynamic range and all, to things like ambient light interfering with contrast ratio, to display non-uniformity). I argue, that this decade is a good time for the industry to begin considering going to MPRT standard (even if old test patterns have to be recycled for now).

To keep things simple, makers of blue-ray test patterns can also list MPRT numbers alongside the "lines density" numbers. That's the simple way without needing to invent new patterns (we still have the capping-out disadvantage, but at least we're now not locked to a specific test pattern)
Okay, makes better sense? Now, moving on:

MPRT is measurable both electronically and by human vision. That's what makes it so promising, like a standardized way -- a universal motion blur number measurement like a contrast ratio measurement (as subjective as contrast ratio measurements can be, with all its complications I've already explained) There are superior ways to measure MPRT than using a "lines of motion resolution" test pattern. That's what some of my discussions in this thread, has been about.

Hopefully this de-mystifies MPRT a little bit. smile.gif

Basically this thread is simply talking about standardizing motion blur measurement.
To the point where it's far easier to get subjective measurements more closely agreeing with objective measurements.
With no bias to a specific display technology.

Thanks,
Mark Rejhon

www.BlurBusters.com

BlurBusters Blog -- Eliminating Motion Blur by 90%+ on LCD for games and computers

Rooting for upcoming low-persistence rolling-scan OLEDs too!

Mark Rejhon is offline  
Reply Display Calibration

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off