Moving camera platform for Motion Blur Measurement -- Any available? ($ for homemade solution)

Some display manufacturers, measuring motion blur scientifically, often use moving cameras (e.g. moving mirror, rotating camera, horizontally moving camera, etc).

Are these devices available pre-built for purchase in the 3-figure range? (Not 4 figures and above).
We need an inexpensive moving-camera motion-blur-measurement device that's accessible to bloggers (e.g. the type of league like myself, AnandTech, HardOCP, TomsHardware, etc).
This enables more objective measurements of eye-tracking-based motion blur (not just subjective)

Alternatively, I am looking for collaborators on building an inexpensive home-made precision-controlled moving camera rail powered by an Arduino. There is $ in it for you. I've posted this in the Arduino Forum. This will also be used to scientifically compare LCD / Plasma / DLP / CRT / etc. in their motion blur, ghosting trails, response-time-acceleration artifacts, phosphor decay trails, pixel persistence trails, etc. It will be also extensively used to test my zero-motion-blur LCD desktop gaming monitor (mod), too. The moving camera will accurately capture all of these trailing effects.

Note: If reviewers (Home theater magazines) and users of motion test pattern makers (Joe Kane, DisplayMate, etc) someday begin using this moving-camera device to measure actual measured "Motion Equivalence Factors" (actual measured version of "CMR" "XR" "Motionflow" "SPS" "FFD" "subfield refresh", etc) -- if you build the first "affordable, accessible, open source" device, your name may be getting publicity for a device that may be eventually used by dozens of reviewers in five years from now, bringing a formerly expensive piece of factory measuring equipment, to the maker revolution and affordable Arduino construction, affordable to future bloggers including myself.
Edited by Mark Rejhon - 2/22/13 at 3:21pm

Addendum: That was fast!
One application by an electronics engineer has now been received already.
I am accepting proposals from anybody until December 7th, and want to compare all the qualified ideas before money leaves my pocket.

Yes, good, I will work to make this compatible with test patterns used for the ADPC method.
That said, it will also be compatible with chase tests (e.g. PixPerAn chase test).
The same test patterns can allow measurements both by human eyes, and by cameras.
Cameras are necessary to "show off proof" in blogs, magazines, reviews. The rigs should now be inexpensive enough to be accessible for blogs.

I will be bringing my rig into a supportive store, and testing a few displays, and publishing the world's first blogger-created actual measured motion blur comparision charts -- complete with individual moving-camera photographs for each display. This stuff was formerly only accessible to larger organizations, but with my Arduino and programming skills, as well as my intimate knowledge of motion blur, I'm trailblazing here. Separately (for a separate page I plan to create) I also have a brand new Casio Exilim EX-ZR200 which has a 480fps / 1000fps mode, which allows me to capture scanning backlight patterns (At 480fps, it is a convenient 8 video snapshots per 60Hz frame), which I'll incorporate as well. Possibly also a few plasma subfield patterns (small slices in macro mode). Will be amazing reading for a niche audience.

Separately, for the motion blur measurement standardization, I currently have a proposal I'm already fishing out to people including Joe Kane (a friend of mine) and Ray Soneria (DisplayMate), since I want to also make sure that their test patterns are compatible with moving-camera-rigs.
Edited by Mark Rejhon - 11/25/12 at 3:28pm

Exactly: I want to measure eye-tracking motion blur.
This moving-camera test is already done by manufacturers. (xrox, would you please chime in about standardized manufacturer tracking camera tests, for our fellow readers who is not familiar with the difference between stationary-camera tests and moving-camera tests?)
I want to do it in a sub-$1000 setup, accessible to bloggers and magazine reviewers.

These two tests (stationary camera and moving camera) measure completely different things.
Stationary camera measure things like pixel-persistence effects, and other factors.
Moving camera measure eye-tracking-based motion blur (e.g. CRT vs LCD vs. LCD strobed/scanned)

An explanation that may help you understand better:
Let's talk about the CRT example. Moving camera tests even at 1/10sec will yield a perfectly sharp image on a CRT -- because a moving camera on a moving image on a CRT display will have almost no motion blur (only 1ms phosphor decay). A camera at 1/60sec will be as exactly sharp as at 1/10sec, assuming perfect object tracking: It's just capturing repeated stationary flickers. (To the moving camera, the moving object is stationary to it). Think of the CRT as a strobe light blinking 6 times if the exposure is set to 1/10sec.

We all know that shorter impulse lengths means less motion blur to the human eye.
*exactly* the same thing happens with camera exposures, too.

When the camera is tracking accurately, it is all relative:
-> To a moving camera, the moving object on the screen is stationary. <-


Given a test pattern moving 960 pixels per second, we have 16 pixels of step per frame. If the pixels were 0.25mm on the display (typical computer monitor), the camera needs to be programmed to move smoothly 240 millimeters in 1 second, to accurately track an object that is moving horizontally 960 pixels per second. Now, testing this on different displays, results in the following:

If it was CRT of 1/1000sec impulses (1ms phosphor), the motion blur is tiny. (~1 pixel of blur at 960pixels/sec)
If it was 120fps@120Hz LCD display, the motion blur is bigger. (~8 pixels of blur at 960pixels/sec)
If it was 60fps@60Hz LCD display, the motion blur is even bigger. (~16 pixels of blur at 960pixels/sec)
If it was a theoretical 1000fps@1000Hz LCD display, motion blur will be tiny. (~1 pixel of blur at 960pixels/sec)
If it was a 1/120sec strobed backlight on a 60Hz LCD, motion blur will be halved (~8 pixels of blur at 960 pixels/sec)
If it was a 1/240sec strobed backlight on a 60Hz LCD, motion blur will be quarter (~4 pixels of blur at 960 pixels/sec)
This of course, assumes, that there's no pixel persistence effects that leak beyond the current frame, and no other effects such as backlight diffusion etc.
(Which is true for most technologies, except older LCD which had a persistence vastly bigger than a refresh)
Of course, real-world would probably be a bit (or lot) worse than many of the above, to varying extents.
Real-world measured motion blur is the honest motion blur -- it accurately measures the amount of motion blur that human eyes sees.

Yes, stationary image test are useful, too. However, moving camera tests are useful for different reasons!
Eye-tracking-based motion blur cannot be measured using stationary camera imaging. It's impossible.
Edited by Mark Rejhon - 11/29/12 at 11:34am

Here are two pictures. One stationary, one moving camera. Taken on an old Samsung 245BW computer monitor with a CCFL backlight (180Hz PWM)


Stationary Camera Advantages:
- Captures pixel persistence effects
- Resolution measurements of best-case resolution scenarios.
- Reference image to compare to moving camera images.


Moving Camera Advantages:
- Capture tracking-based motion blur (in additon to any pixel persistence effects)
- WYSIWYG as seen by human eye (eye-tracking blur equals camera-tracking blur) (scientifically confirmed)
- Captures double-image and triple-image effects (triple image at 180 Hz PWM for 60 Hz refresh) (visible in photo above!)

Taken from my makeshift manual moving camera rig, which I wish to replace with precision electronically-controlled camera tracking.
P.S. Camera tracking (1/10sec shutter) will also capture the double-image effect of 30fps @ 60Hz (on impulsed displays). It's WYSIWYG; scientific record of what the human eye saw. It does not capture fine-level temporal effects (e.g. subfield dithering) but it captures a lot more of what the human eye saw, than stationary cameras.

In the moving image, there's approximately 15 pixels of motion blur on both sides of the UFO graphic; about accurate for a 960 pixels/sec moving object (16 pixel jump per frame at 60Hz on a sample-and-hold LCD). If I repeat this test on a 120Hz sample-and-hold display (that has no major pixel persistence effect beyond one refresh), there will be half as much motion blur (~8 pixels of blur).

See?
A picture is worth a thousand words.
Edited by Mark Rejhon - 11/28/12 at 9:38am

Update -- got two more offers privately (through my post in the Arduino forum) to build this moving-camera rig. Price quote (for parts and labour) is currently within my budget, hundreds of times cheaper than expensive corporate setups...

The favourite approach is an inexpensive cogwheel-drive with motor speed monitoring, because that's what inkjets use. They use a cogwheel belt and a cogwheel and are able to precisely spray a dot at micrometer-league precision while the ink cartridge is moving at 1 meters per second. Very impressive technical engineering achievement and refinements over the years; so no need to reinvent the wheel. The challenge is variable speed control and maintaining an accurate speed after accelerating the camera's mass. Position of camera and timing of shutter button doesn't matter (shutter lag is okay). What is important is that the camera is moving precisely smoothly at a precise speed (within required accuracy specifications) for long enough, regardless of whether the shutter clicks early or late in the smooth-movement window. Bonus, if the motor rotation data can be streamed too, so that it can be translated in software to approximate linear position of the camera, but this is not required.
Edited by Mark Rejhon - 11/28/12 at 10:48am

Yep, my device is essentially a far-higher-speed version of a dolly slider, but for live motion blur capture, not for time lapse.

The chief problem is that it needs to move accurately at a fast speed (250mm/sec, preferred accuracy +/- 1mm/sec, and configurable adjustable in 1mm/sec increments). The existing dolly solutions are unable to achieve this precision. Meaning, I need to take a picture while the camera is moving exactly, say, 227mm/sec or 245mm/sec (adjustable in 1mm/sec increments), and the motion of the moving camera doesn't vary more than 1mm/sec while taking a picture. That allows me to test any kind of display, and program the moving camera based on the dot pitch of the display, so that the camera scrolls at, say, 960 pixels per second (+/- a tiny fraction of a single pixel over a single 1/60 period, and still +/- a fraction of a single pixel at 1/10sec too). Since the primary displays for testing this rig on are 24-27" displays, that requires approx 250mm/sec to follow a moving object moving at 960 pixels per second. Later, I'd appreciate compatibility with up to 60" displays (even if I have to track moving objects moving only at 480 pixels per second).

The camera must be moving steadily and accurately while the shutter is open for long periods (e.g. 1/60sec, 1/30sec, 1/10sec) for live motion blur capture. That is, I need the camera to move almost as fast as an inkjet cartridge head, moving WHILE taking a picture. (The shutter trigger button can be modified and attached to an Arduino circuit or other microcontroller). The dolly's cant do it that fast while having accurate speed during the time the shutter is open. (To make the project easier, I have no precision requirements for the acceleration phase or positioning -- just that the speed is as close as possible to my preset, and the motion is as smooth as possible, while the shutter is open.

A custom modification to an existing dolly slider, to be able to pull this off, is an allowable proposal I'd be willing to pay someone else labour and parts to do. But the dolly cannot cost too much (many cost 4 figures) -- ultimately, at the end of the day, it needs to be the world's most inexpensive tracking-camera motion blur capture-and-measurement system, accessible to a blogger's or magazine reviewer's budget. This is the kind of stuff normally accessible only to large corporations. But I've at least determined the technology is doable on the cheap, thanks to commodity inkjet technology. I have gotten price quotes half as big as the price of a complete prebuilt dolly not designed for active motion blur capture (So far, the price quotes I got, are about ~$500 labour + costs of parts, which is within my budget)
Edited by Mark Rejhon - 11/28/12 at 3:49pm

About overlapping frames: This *specific* camera image is only 1/60sec, so the blur trail is 3 copies caused by the 180Hz PWM CCFL backlight -- it strobes 3 times per refresh. The triple-image effect is visible with the human eye. But, yes, it's essentially tantamount to 3 repeated frames. The picture is WYSIWYG -- what I saw with my human eye: Triple image during fast movement.

About measurement I plan to be using different test patterns, such as a moving grid, which allows me to compare the ratio of thickness of vertical details, to the thickness of horizontal details. The more blur, the more the ratio distorts. Things like this, will allow easy blur trail measurements. However, I'm not yet at liberty to announce specific measurement details. Keep tuned!

On a related topic, I just recently became an associate member of Society for Information Display (www.sid.org), so over the coming months (years?) I'll get familiarized with various professional industry standards of measuring motion blur; and attempt to develop an open-source, blogger/magazine reviewer-accessible method of active tracking-camera tests. It probably will take me a year of hobbyist spare-time or two to trailblaze (and package it in a user-friendly manner, like a catalog of comparable motion blur ratings and photos), but this is uncharted territory at the bloggers/magazine reviewer league. There are many standards like APDC, MPRT, etc, but these will need to be adapted to meet simplicity requirements for Average Readers (like family members, friends, etc)

Here's a paper on moving-camera tests ("pursuit camera"):
http://144.206.159.178/ft/CONF/16408531/16408542.pdf
However, this is advanced reading.
Edited by Mark Rejhon - 11/29/12 at 11:27am

More information on part of the purpose of this tracking camera I'd like to get built inexpensively.
In the long term, I am working on a new motion benchmarking standard (M.E.R. = Motion Equivalence Ratio) that allows people to compare displays based on actual-measured numbers, instead of manufacturer-claimed numbers.

The honest "Measured Motion Equivalence Ratio" is an actual measured version of the numbers found in industry lingo (e.g. Samsung CMR 960, Panasonic 2500 Hz FFD, Sony Motionflow XR 480, plasma 600 Hz subfield drive, etc). For example, Samsung CMR 960 might actually only have a MER 320, displays with 240 Scenes Per Second may actually only have a MER of 180, and a Pansonic 2500Hz FFD might only have a MER of 200 (due to 5ms phosphor decay), etc. CRT's will have extremely high MER's, naturally (CRT becomes the benchmark to beat in motion clarity improvements for future display technologies). Measured Motion Ratios are relative to the "ideal reference 60Hz sample-and-hold display" (which would be MER 60) similiar to the following. Much like measured contrast ratio is always lower than laboratory/claimed contrast ratio.

Concept (mock-up) user-friendly graph below:

(concept future magazine-reviewer graph comparision of motion blur.
Bigger bars means less motion blur = shorter measured blur trail.

Formula is very simple:
Motion Equivalence Ratio (MER) equals the the pixels per second, divided by the length of blur trail

960 pixels per second motion with 16 pixels of motion blur trail = MER of 60
960 pixels per second motion with 8 pixels of motion blur trail = MER of 120
1500 pixels per second motion with 10 pixels of motion blur trail = MER of 150
480 pixels per second motion with 3 pixels of motion blur trail = MER of 160

This is part of my purpose of designing a tracking-camera rig that costs less than $1000 to create: Within blogger/magazine budgets, rather than a niche kept in TV manufacturer leagues, big university labs, and large budget research organizations. The equipment already exists, papers are already written, but I want to bring it to the bloggier/magazine level & make it possible for people to create user-friendly graphs like this. Attractive to various niche audiences, such as people who play video games at 60fps and want the least amount of motion blur, etc.

The motion benchmark standard, I am developing, M.E.R. has many advantages of being a motion blur measurement standard that can be directly compared between LCD and non-LCD technologies, compared between different resolutions (e.g. it's possible to benchmark a 1080p display against a 4K display), and can be done using moving test patterns that is measured either by human eyes (PixPerAn test pattern compatible too -- chase test -- and compatible software such as DisplayMate Motion Bitmaps, that can be configured with horizontally scrolling test patterns running at an exact pixel rate). Test patterns similar to the PixPerAn chase test, can be used to calculate an approximate M.E.R. value using only human eyes (no camera). That gives people choice to use a camera (or not) to measure M.E.R. values. In addition, this will be good for the advancement of LCD display engineering too, as this is also still useful to manufacturers, just packaged in a user-friendly manner using a motion benchmark rating (M.E.R.) that's based on Hz numbers, but in a manner that accurately represents actual visible motion blur seen by human eye (taking into factor of all interpolation and impulse-driving enhancements to motion) --

There are people (like myself) who need _measured_ numbers for motion blur, much like measured contrast ratios versus manufacturer-claimed contrast ratio. A measured number that's display-tech independent, motion-enhancement-technology-independent, interpolation compatible, scanning backlight compatible, plasma/CRT/LCD/DLP compatible, resolution-independent (allows comparing 1080p vs 4K displays in motion blur), takes into account of *everything* that causes motion blur (not just pixel persistence, not just phosphor decay, not just sample-and-hold, etc), future-proof.

In short: I am developing a blogger/reviewer-accessible benchmark value for motion blur -- easily understood by average readers.
-- an actual "measured" version of the numbers from Samsung CMR ("480" and "960"), Sony XR ("480" and "960"), 240 Scenes Per Second, MotionFlow, 2500Hz FFD, 1600Hz scanning backlight, etc. My number will completely unify all of that into a single, honest, actual seen-by-eye motion blur measurement number. Although it is not necessary to use a camera to calculate Motion Equivalence Ratio (MER), it substantially improves accuracy of MER calculations & provides article-friendly blur comparision photos between different displays (convenient!)

This is a related goal -- The tracking camera (that I want to get built) is also to test the scanning backlight I'm building, and to compare it to other displays.
Edited by Mark Rejhon - 11/29/12 at 11:41am

[Had a bit of a crash course reading up on ADPC and MPRT] xrox, thanks for explaining, and thanks for pointing out the precedents of tracking-camera tests.
This is an excellent start to motion resolution testing in the last few years, though the motion test pattern should be measured differently for everyday consumers -- "X lines of motion resolution" have been a quote I've read in reviews. I feel this is mostly meaningless to Average Consumers when it's tied to a specific benchmark that's (on average) hard to obtain...

FPD benchmark disc and sinebursts are a good start, but motion resolution measurements measured from that disc is generally not directly comparable to motion resolution measurements made through other methods (e.g. non-APDC patterns, etc). Also FPD is a disc created by a specific group (plasma), although it is usable for all other technologies. Various factors makes standardized motion resolution comparisions difficult, if you're tied motion resolution benchmarking to just one disc.

I believe that a simpler motion resolution measurement standard, "Measured Motion Equivalence Ratio" (MER) is much simpler -- "MER 240" would be equal to a scientifically ideal 240fps@240Hz on sample-and-hold display, or scientifically ideal 1/240sec impulses on a impulse-driven display. And it scales better (beyond "1080 lines of motion resolution"), and is resolution-independent (while sinebursts are more resolution-dependent). People are already familiar with refresh rate, and "Hz" - with all the manufacturer hype (600 Hz subfield, Clear Motion Ratio 960, etc), so the MER (as a measured motion blur equivalent of such numbers), is more useful and display/resolution independent. Even only testing a few different motion tests, I've found that many different kinds of motion test patterns can be used to calculate an MER value. WIth MER, anybody can invent proprietary or standard moving test patterns (disc based or software application based) that yield a standardized MER value that can be compared between different displays, regardless of technology and resolution.

I am putting together a standards document (a very rough draft) and contacting some key people at the moment, to try and trailblaze such a simpler standard more accessible to mainstream. I just recently obtained a membership to Society for Information Display, to help assist in this matter, and in researching my options. I've got experience writing standards documents (Such as my XMPP Extension XEP-0301 for an instant messaging feature on XMPP networks, over 400 hours spent on this document over the last 2 years through standards collaboration.) so my existing standards-document experience will help me along here. There's been a lot of work done before me, but a lot of it is too complex, or requires a very specific test, etc. The MER standard that I've come up with is a more generic method of measuring motion blur -- MER is just simply pixels-per-second of motion movement, divided by pixels-of-motion-blur-trail. Completely independent of test pattern, completely independent of display resolution, future proof motion resolution measurement standard.

Meanwhile, I've been trying to obtain a copy of the FPD benchmark disc [or original-high-bitrate video files, if they're legally available publicly] -- to determine if it can also be used for the "Measured Motion Equivalence Ratio" method of measurement. I want to test all kinds of motion discs to see how their motion test patterns can be converted into a MER value. It takes only a quick 5 minute PixPerAn (chase benchmark) to quickly calculate a MER value for your display, and I'd like to find out if FPD benchmark disc, and whether existing APDC patterns, are also usable to calculate a MER value too. Where can I get an FPD disc? Also, similar discs such as this one is out of print, and they don't show up on eBay, so obviously, these test patterns aren't that easy to obtain. Fortunately, my MER is measurable from more types of test patterns than just a very specific APDC-pattern, making motion-blur-measurements more accessible. (This does not preclude ability to create proprietary test patterns to earn money off from -- people can invent better mousetraps -- such as better/more accurate MER-measurements from a proprietary test patterns that allows human eye to measure MER more accurately than existing tests such as PixPerAn chase test.)

In addition, are existing APDC patterns used to calculate exact MPRT values, too? If so, that's good news for MER, because MER is actually the inverse of MPRT (MER = 1000 / MPRT value), though the honest MPRT eqiuvalents needs to account for the impulse-drive (so a 2ms 60Hz LCD actually should really be a MPRT of 16ms due to sample-and-hold effect, and 1000/16ms = a MER of 60). The problem with honest MPRT measurements like those -- "16ms" -- gets confused too often with pixel persistence -- "2ms" -- by consumers, so it is not really a useful benchmark for comparing products for Joe Consumer. Thus, the MER standard I am developing, is a lot more useful for everyday mainstream blog readers, because MER numbers are universal; CRT/plasma/LCD/DLP/etc compatible -- and easy to compare to the refresh-rate-equivalence numbers quoted by manufacturers.

Also, 6.5 pixels per frame suggested in the APDC standard is way too slow for future display technologies (I hope that they have much faster rates available); my strobed-backlight display (250W in a 24" LCD) will be remaining fully sharp at 16ppf -- more than twice as fast as the APDC pattern. My motion benchmark will be more realistic for video game players, and first-person shooter videogames utilize faster motion than 6.5 pixels per frame, and some trained-eye videogamers can still detect extremely minor motion blur even on CRT too (phosphor ghosting effect) -- that requires at least 20-30 pixels per frame in order to be easily detectable. Also, MER (the number I am developing) can be measured from a moving test pattern running at any speed that the human eye can track at. Video games are a more aggressive test of motion blur than a lot of digital video (e.g. slower shutter speeds masking display motion blur, over-compression artifacts adding blurriness that masks display motion blur, etc), and I am a video gamer, and today's MPRT/APDC tests has generally been inadequate for this audience.

MPRT and ADPC numbers are good and useful for advanced readers and display industry professionals, but are mostly meaningless to everyday readers (my family members and my friends), people keep asking "is that a 240 Hz display?" when they don't really know two different 240Hz displays may actually have very different MER's (e.g. one of them may actually have a MER closer to 140 and another has a MER closer to 200). People can understand MER better when it's a honest number directly compared to "Hz" and "refresh rate" numbers, directly compared to the numbers shown at Best Buy ("CMR 960" "Sony XR 480", "SPS 240", etc...) Even my friends/family likes the idea of a "honest measured verification" number that can be directly compared to manufacturer claimed numbers of "120" "240" "480" "960" "600" "1600" "2500" etc... That's what the Measured Motion Equivalence Ratio (MER) is -- a motion benchmark number accessible to everyday readers.

(P.S. I am looking for additional peer reviewers for my MER standardization. I have an early draft document in progress, that I may eventually submit to a standards organization within a year. Or if you know of anyone standardizing a MER-like number, please refer me to them. Contact me by PM or email.)
Any that sells such systems for $1000 or less? Most systems I've found are beyond the range of blogger budgets, etc. I've found a builder quote that puts me in the ballpark.
Edited by Mark Rejhon - 11/29/12 at 10:37pm

Doing some more MPRT research, and reading some SID.org papers....

I will probably piggyback off the existing MPRT standard -- why reinvent the wheel.
And, of course, MPRT tests are done by pursuit cameras (exactly the same kind of camera I'm trying to get developed for cheap)

The good part is that, the papers (which I have paid access to now as a SID member) - "Moving-Picture Response Time and Perceived Motion Blur on LCD Panels" (registration required for full PDF) -- Very useful quotes that confirms what I've always known:
"...it has been confirmed that the MPRT method is a measurement method that can quantify motion blurs on LCD as perceived by humans..."
"...To digitalize such motion blurs, we need a measuring method that takes eyeball motion into account..." (tracking cameras)
There are also commercial motion-blur tracking cameras like these, but are out of my budget range for an independent blogger like myself.
(That's why I've posted this thread as a call for an electronics hobbyists to build me one.)

Obviously, the existing MPRT research I'm reading is quite handy.
I've confirmed that any MPRT-compatible motion tests are conveniently directly convertible to a "Motion Equivalence Ratio" (MER) simply by using the formula (1000 / MPRT)
Edited by Mark Rejhon - 11/29/12 at 10:18pm

Joe -- good reply, good questions. Actually, this will work. That's what some commercial products do.

In fact, I plan to do exactly this with a Casio Exilim EX-FC200 which I just ordered directly from Japan for only $300. It is a consumer camera that has a low resolution high speed 1000fps mode (224x64). However, that's enough resolution, when used in macro mode, to determine motion blur over a few pixels (say, 1 LCD pixel to about 2 or 4 camera pixels, or about 50 to 100 LCD pixels at close-focus range, and captured at high speed). A 960 pixels per second moving object would have 16 pixels of step, so I'd accurately capture edge blurring from the samples -- all I need is to macro-focus close enough to capture a moving edge. I'd horizontally frameshift each frame using a computer program (relative to the speed of motion I've displayed on-screen), and stack the frames to get the virtualized motion blur image. This would allow me to measure MPRT to a 1ms accuracy. The 1000 is not divisible by the 60 Hz, so I will get some framerate aliasing, so the accuracy range may actually be more like 2ms (a span of +/- 1ms of the measured value).

However, the Exilim 1000fps method, at the cost I've paid for -- there's no 1000fps cameras commercially available with megapixel resolution at 1000fps for under ~$5000(ish), and I've got to live with thumbnail sized 224x64 videos, purely for MPRT measurements, but may not be usable for actual photographs of a moving object onscreen. That's why I'd prefer a pursuit camera rig, so that I can have actual high-resolution pictures of moving objects to "show off" in my blog, from my tests.

Nontheless, the approach that you suggested (that I was already aware of!), is only $300 plus open source computer software, this could be the world's lowest-cost scientific method of measuring MPRT via a "tracking" technique, using scientific equipment, though it would be limited to +/- 1ms precision. This is less than 1/10th the cost of commercial MPRT-measuring (pursuit camera) setups. On the other hand, if you don't mind doing subjective measurement by human eye, PixPerAn allows you to manually measure MPRT via the chase test, to a roughly similar accuracy, using just your human eyes. Apparently, it's fairly accurate.

I was hoping to achieve sub-1ms precision, so that I can eventually measure MPRT's as low as 0.2 millisecond (equivalent to MER 5000, or a theoretical 5000fps@5000Hz, or a CRT with theoretical ultrashort-persistence phosphor), because my Arduino controller for my home-made backlight will be able to make the flashes as short as 1/10,000second. I want to be able to test all the way to the complete disappearance of the "point-of-diminishing" returns -- aka accomplishing a zero-motion-blur LCD. I won't be able to achieve that with only a 1ms MPRT precision, since a 1000fps camera is limited to a MPRT of only 1ms.

So, yes, this method would work -- with a sufficiently high-framerate camera (1000fps), and it does have its pros (lack of vibration, and very cheap cost for MPRT +/-1ms) but it has many cons (no ability to measure MPRT more accurate than 1ms precision, poor images for magazine/blogs). At one point or another, I might use both methods concurrently as checks on each other, to verify the accuracy of the pursuit camera. Yes, but the motion blur has been shown scientifically the same as rotating (rotating a camera or rotating an eyeball). It's much harder for a hobbyist/blogger to calculate the correct speed of rotating the camera (or rotating a mirror that the camera is pointed to). The rotation speed is non-linear as the moving object passes by, since the screen isn't curved around the camera's point of rotation. So more difficult to keep it all synchronized, provided I'm moving the camera slow enough that vibrations are not yet a limiting factor. Sliding the camera is far, far easier to mathematically calculate, for a hobbyist-type setup. I suspect that vibrations will be the limiting factor, probably, though I am working on figuring that out, and I don't think they'll be a major problem at 1/60sec exposures while moving at ~250mm/sec (the speed needed to photograph a 960pixels/sec object on a 24" monitor). Multiple stationary camera will work but the prerequisites suddenly becomes: Exposures must be shorter than the MPRT accuracy. If you're trying to measures MPRT within 1ms, you need 1/1000sec exposures. If you're trying to measure MPRT within 0.2ms, you need exposures of 1/5000sec, etc. Suddenly, it doesn't make economic sense anymore. To capture a whole refresh (16ms) at MPRT +/-0.2ms accuracy, you need to pursuit the frames every 0.2ms (either via moving camera, or via stacked stationary images). Via stacked stationary images, you would need to take a picture every 0.2ms for a whole 16ms. That's a whopping 80 photographs, and that's a whopping 80 cameras. Even at $80 per camera, that's still $1600 worth of cameras, and it will be very hard to point all 80 lens at the same location on the screen.

You need the camera to continuously collect photons for the entire refresh, and this is easiest with a single-camera pursuit camera setup, or with a sufficiently-high-framerate camera. (e.g. 5000fps camera for measuring MPRT +/- 0.2ms). You can kind of "cheat" this requirement a little bit, by stopping taking pictures when the picture is completely black (e.g. after CRT phosphor is about 90%+ decayed, or backlight is turned off and LED phosphor has decayed), so theoretically you could capture just the strobe, and stop capturing the dark moments between frames, as long as you've accurately timed how much dark time there is between the strobes. But to be truly universal to all kinds of displays (regardless of how it "smoothes" motion -- via interpolation or via flicker) -- I really need to capture continuously for the whole refresh cycle.

(NOTE: I'll probably quote my MER numbers (which is calculatable via 1000 / MPRT), since it's easier to compare MER numbers to the jargon (CMR 960, SPS 240, Motionflow XR 480, etc), as MER is essentially a honest measured equivalent thereof, of such manufacturer claims)
Edited by Mark Rejhon - 12/2/12 at 3:14pm

[some boring math]
BTW, my accuracy requirement of 250mm/sec +/-1mm accuracy, allows tracking an on-screen object at 960 pixels per second on a 24" computer monitor. A 24" monitor is approximately 500mm wide, so the pursuit takes 2 seconds to transverse a 1920x1080 pixel screen. Approximately 0.25mm per pixel. Between refreshes at 960 pixels per second, is a 16 pixel step at 60Hz. At 0.25mm per pixel, that's 4mm of movement. +/-1pixel at 250mm/sec translates to +/-1micron at 0.25mm, so that would be an error of +/-16 microns (0.016mm) during a 1/60second photograph. Here, camera vibrations will probably be the limiting factor. But 0.016mm accuracy is more than an order of magnitude over 0.25mm for a pixel, that's an order of magnitude more accurate than needed to measure MPRT at the full pixel level. This might or might not be achievable, and may have live with more relaxed accuracy targets.

However... as the camera will be in macro mode to make the monitor pixels big, I'll be able to measure sub-pixel motion blur at this level, without needing a pursuit camera for faster movement (e.g. 1920 pixels per second or 3840 pixels per second) which is extremely hard to do. For 0.016mm accuracy for 0.25mm pixels, for objects moving at 960 pixels per second, that allows MPRT of accuracy (1/960) / (0.25 / 0.016) = within 0.1ms accuracy. More than ten times more accurate than the 1000fps camera method. And I'd only need to macro to the point that monitor pixels are spread over at least 10 pixels on the camera's CCD, in order for the MPRT +/-0.1ms accuracy calculation to be valid.

If this is not possible to do with a moving-camera rig (I'm still discussing; consider inkjet printers are able to spray dots at micron-league accuracy while moving a cartridge-array at almost a full meter per second! That's why I recommended an inkjet mechanism), then I'll have to settle for MPRT +/- 1ms using the Casio Exilim EX-FC200 method of cheap 1000fps for the MPRT measurements.

Obviously, I'd need to test on a reference (e.g. multiple passes past my homemade LED strobe flashing at 1/10,000sec) to make sure that the camera vibrations isn't blurring images at 1/10,000sec flash, in front of some kind of a fine-print test pattern (or using my scanning backlight setup). A pursuit camera meeting my required accuracy, tracking an object at 960 pixels per second on screen, moving past my backlight strobing at 1/10,000sec (to be confirmed by oscilloscope+photodiode to make sure LED phosphor decay isn't limiting factor) -- would blur pixels less than 1/10 pixel width. This means a camera moving at 250mm/sec that's moving at 0.016mm accuracy during 1/60sec, would not be able to blur the red, green, blue subpixels (about 0.08mm) -- it'd look almost exactly like a stationary camera image even while the pursuit camera is moving at 250mm/sec -- no motion blur.
The image will probably be fairly dim at 1/10,000sec even with my 250 watt backlight (do you know how bright even just a small 30-40 watt section looks like? I've posted photos at www.blurbusters.com in an earlier blog entry).

Either way, I really want to be able to measure MPRT down to +/- 0.1ms, so I can scientifically measure the "zero motion blur LCD" -- less motion blur than CRT. I can't do that with MPRT setups only accurate to +/-1ms. Another option of mine is renting someone else's 5-figure priced MPRT setup (university, etc) and getting them to measure my display.
[/some boring math]
Edited by Mark Rejhon - 2/22/13 at 3:21pm

I have a breakthrough. Keep tuned at www.blurbusters.com this spring.

Homebrew solution. No Arduino necessary (optional). Can take advantage of any digital camera with adjustable exposure. (no major camera restrictions). Not very scientifically accurate, but very blogger-friendly, and far more accurate than common PixPerAn photo captures (found in places like prad.de, tftcentral.co.uk, pcmonitors.info, etc)
Edited by Mark Rejhon - 3/11/13 at 11:29am

Blur Busters became the world's first blog to use a pursuit camera.
I have posted the first photographs from the pursuit camera. It is working!!
There are scientific test patterns that I can use (for measurements), and further accuracy improvements, but my Blog focuses on the Plain English; that's what I aim for.

Next exercise over the next few months: Run the pursuit camera at different refresh rates, CRT, Plasma, LightBoost.
(Once I launch the new Blur Busters Motion Tests designed for computer users/bloggers/reviewers)
It includes very accurate MPRT test patterns usable by everyday computer users (+/- 0.1ms to an oscilloscope photodiode)

Sample pursuit camera photography at:
LCD Motion Artifacts 101

True that it is a niche market thing, it's more mainstream of a niche market than you think:
Why does 120 Hz and 240 Hz HDTV's exist?
Motionflow, Clear Motion Ratio, Scenes Per Second, etc.
Europeans love their 100 Hz and 200 Hz TV's, too.

One problem is almost none of these modes are video-game compatible or computer friendly, due to input lag of interpolation, and being disabled in Game Mode (with a few exceptions). Blur Busters Blog focusses on computer-compatible and game-compatible motion blur elimination technologies.

Another problem is people don't know it until they see it. People don't notice 3:2 pulldown until they are taught to look for it. People don't recognize 30fps vs 60fps until they see a side-by-side comparison. People don't notice DLP rainbows until they look for it. People don't see videophile picture quality until they're taught about it's attributes. Blur Busters Blog aims to educate, one small group of people at a time, ever since it started less than half a year ago. The blog now gets over 10,000 pageviews a week, with thousands of unique visitors and it's still ramping up as time passes (no ads yet, either, as of March 2013.). The HardForum thread has over 75,000 pageviews with over 1,000 posts. The overclock.net thread is not far behind. Random people are now suddenly talking about LightBoost, many dozen English forums and filtering into even international forums (e.g. russia, germany) without my help; because the news is spreading between people in the high-end computer gaming communities and slowly filtering down to the mainstream "wow, I should try it" forum audience in those far flung locations.

A feature just needs to be sufficiently "popular enough" to become standard. 24p was a luxury until it became standard in both players and HDTV's. TiVo PVR's were luxury machines until cable companies made them standard. 120 Hz TV's were luxury until motion interpolation became standard in most HDTV's. Blur Busters Blog has already influenced one monitor manufacturers to compliment on my work, doing a better job than nVidia in selling LightBoost(tm) monitors in some channels. (LightBoost is an nVidia trade name licensed to monitor makers, for 3D glasses usage; but apparently also benefits 2D zero motion blur). This may influence some standardization of more widespread video-game-compatible and computer-compatible motion-blur elimination technology (after earlier failures from poor attempts). I'm putting money in manufacturer's pocket today.

Google "LightBoost". Look at the top two hits.
The LightBoost zero motion blur capability is already more popular than for its original intended "3D glasses" purpose.
Edited by Mark Rejhon - 3/13/13 at 8:05pm