Setting Contrast - why above white does not clip - Page 2 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 2Likes
Reply
 
Thread Tools
post #31 of 165 Old 05-20-2009, 01:00 PM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Legal, schmegal.

His statement and my point are about valid and invalid. The quoted statement made no reference to 'legal' anything. I think valid and invalid are pretty well understood terms and concepts and you're introducing legal here as a red herring.

One theoretical advantage I can see with sRGB is there is no concept, room, allowances or reservations made for any INVALID and OUT-OF-GAMUT 'data' or errors. (BTB/WTW) It doesn't reserve 36% of it's codeword volume for such INVALID data or errors. sRGB uses all of it's 8 bits to represent VALID, IN-GAMUT data.

Which seems better to you- A 8 bit color space with 16.7 million valid, in-gamut codewords (sRGB) or one with 10.6 million valid, in-gamut codewords (StudioRGB)?

Dave
dlarsen is offline  
Sponsored Links
Advertisement
 
post #32 of 165 Old 05-20-2009, 01:24 PM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I note that in a few of the images Stacy has posted, there are many BTB pixels outside of the active image area and well into the encoded black bars area. How could these pixels possibly even be a candidate for valid image information? If we agree that the encoded black bars should be black, then all these BTB pixels that are out of the active frame area can ONLY be considered as errors. They should be 16,16,16 and they aren't.

If this was transformed and rendered to sRGB levels, these errors in the encoded black bars would be corrected.

Also, if it's argued to be a valid extension of image information, why the need to invert (and destroy) the very data in question? Why not post it otherwise unaltered in A/B manner such that it can be contrasted and compared?

Dave
dlarsen is offline  
post #33 of 165 Old 05-20-2009, 01:27 PM
AVS Forum Special Member
 
sspears's Avatar
 
Join Date: Feb 1999
Location: Sammamish, WA, USA
Posts: 5,320
Mentioned: 23 Post(s)
Tagged: 0 Thread(s)
Quoted: 315 Post(s)
Liked: 512
Quote:


Let's not forget that editing is done in RGB space... probably 14- or 16-bit, though it's possible there's some 18-bit equipment out there today that I don't know about. And it is EASY to constrain the editing station software to the 14- or 16-bit equivalent of 16-235 8-bit space or to edit in "full range" space and constrain the data to 16-235 later when the original data is converted from its original 14, 16 or maybe 18-bits to 8-bit 16-235.

There is offline editing and online editing. Off line editing is done on a proxy, which is usually low resolution, etc... This is how a lot of stuff is initially assembled.

Most color grading tools operate in 32-bit floating point. I use SCRATCH for my color grading work. Lustre is another popular tool. SCRATCH is also how I conform an offline edit. I can edit in Final Cut Pro and what the video looks like does not matter because the only thing I get from FCP in the end is an EDL. (Edit Decision List) This EDL lists the clips and mark in/out points and trasition effects.

I know EFilm has several seats of SCRATCH as well as Lustre rooms.

Stacey Spears
Co-Creator, Spears & Munsil UHD HDR Benchmark
sspears is offline  
Sponsored Links
Advertisement
 
post #34 of 165 Old 05-20-2009, 02:05 PM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
I honestly think clipping below reference black is not nearly as big a deal as clipping above reference white. I've seen some content that looks visibly changed when below-reference is clipped, and I assume mostly it's because those lower values affect the scaling and processing algorithms. On a CRT, a really low below-black value has the potential to affect the next pixel, pushing it up just a tad because of filtering, internal ringing, etc. On a fixed-pixel display, that's not necessarily going to happen if you get 1:1 pixel mapping, and I can't honestly say I've seen any visible artifact because of below-reference clipping in a situation like that. I still think there's no good reason to clip below 16; it's not like it's helpful in any way.

But whether or not displays should show the above-reference signal is an easy question. They should. The content was viewed all throughout the video production and mastering process on BVMs that do not clip above reference. And ALL content has some above-reference material. All of it. You can see the difference by switching from clipped to not-clipped. The not-clipped is correct, and the clipped is incorrect.

Some of the above-reference material is basically unimportant. It just doesn't change the image in any visible way. Some of it, on the other hand, is easily visible when you compare clipped to not-clipped. Does the image look basically OK if you clip? Yes, most of the time it does. But that proves nothing. If you clipped at 225 the image would still look OK most of the time. You could clip black at 25 instead of 16 and I bet a lot of viewers would say the image looked punchier. And in fact it would be punchier, with more contrast. But it would still be wrong, and it would still look different from a reference monitor. And making the video look like it does on a reference BVM is the goal.

Bottom line: there are visible differences in the image when you clip above 235. No professional monitor clips above 235. If you want your home system to look as close as possible to a pro monitor you will not clip above 235. Q.E.D.

Don
dmunsil is offline  
post #35 of 165 Old 05-20-2009, 02:16 PM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Oh, and whoever said that above-reference values are out of gamut was wrong; increasing the brightness range of R', G', and B' doesn't move them out of gamut. {235, 16, 16} is the same color as {254, 16, 16}, and is way in the far corner of the gamut triangle. One is just a brighter version of the other. Conversely, there is a visible difference between {235, 235, 235} and {254, 235, 235}, but both colors are fully in gamut. One of the basic color principles is that the color of any all-positive combination of three primary colors lies inside the triangle formed by those primaries.

Now, below-reference values can be interpreted as out of gamut. The new xvYCC standard uses the below-reference space to represent a wider gamut in a mostly backward-compatible way. But in order to be out of gamut, a color needs to have at least one negative component.

Don
dmunsil is offline  
post #36 of 165 Old 05-20-2009, 02:41 PM
Senior Member
 
kamui's Avatar
 
Join Date: Mar 2005
Location: Fort Myers, FL.
Posts: 312
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 38 Post(s)
Liked: 18
Quote:
Originally Posted by dmunsil View Post

I still think there's no good reason to clip below 16; it's not like it's helpful in any way.

Now I'm confused. You "DON'T" want to clip below 16? I thought that you wanted to perserve whiter than white, but clip blacker than blacks.

In other words, you should be able to see visible differences in everything between 235 and 254, but everything from 1 to 16 should be the same "blackness".

What I usually do is take a black 16IRE pattern in 4:3 format and make it match the pillarbox sides of a 16:9 screen, so that 16IRE matches the blackest blacks my monitor can produce. Is this incorrect?

Sorry that I'm going the opposite way of the thread title :P

-Brian

sspears:
(Rules? Rules? In Television???? Thou shalt not oversaturateth red. RGB shall be thy rod and thy staff; there shall be no other color space before thee.)
kamui is offline  
post #37 of 165 Old 05-20-2009, 02:41 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,735
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 26
Quote:
Originally Posted by Doug Blackburn View Post

Let's not forget that editing is done in RGB space... probably 14- or 16-bit, though it's possible there's some 18-bit equipment out there today that I don't know about. And it is EASY to constrain the editing station software to the 14- or 16-bit equivalent of 16-235 8-bit space or to edit in "full range" space and constrain the data to 16-235 later when the original data is converted from its original 14, 16 or maybe 18-bits to 8-bit 16-235.

Sure, that would be easy to do. But that isn't done most of the time, and the proof is in the content that we have. If you look at the values in the content, you can tell whether it's been clipped off or not, and almost anything I've looked at on DVD isn't clipped off at 16-235 in RGB. I don't have the ability to examine BD, but Stacey does, and the images he just provided should be interesting for you to observe. There is no good reason to clip off the content at 235, particularly when the peak white range 236-254 is explicitly provided in Rec709 for video values above 235. So when you have a video standard that provides you room for transients or specular highlights or heavily saturated bright objects above your nominal reference white, what advantages do you get by clipping all that off? That's a rhetorical question.

Quote:


Data doesn't start out as YCbCr... YCbCr is an intermediate step.

Obviously. Going back to square one doesn't really deal with the discussion thus far.

Quote:


YCbCr data originates as edited RGB that has been downconverted and constrained to 16-235 during the downconversion process.

Except that it hasn't been constrained to 16-235 in RGB. There is no good reason to clip the content this way during mastering, and if you actually looked at the content you'd know that it hasn't been.

Quote:


Using your example of 250, 181, 240 (RGB - I'm not checking the conversion math, I'll assume it is correct)... the point defined by RGB 250, 181, 240 will not exist in your original RGB data (before it is converted to YCbCr).

Why wouldn't it? It certainly could. It's a perfectly legitimate piece of video data. Unless a hard clip was applied at 235 (which as evidenced by numerous examples above, and in many threads previous, and from a variety of sources is not common), then that pixel could conceivably exist in a source without any difficulty or the sun spinning off its axis or the disc player exploding in a ball of fire. It happens all the time. Everytime you put a disc in your player you're probably looking at values like that. Unless of course you decide you know better and clip it off.

Quote:


Assuming you are using YCbCr encode and decode matricies that aren't screwed up, you will never encode/create a YCbCr triplet that would decode to an out-of-gamut RGB value.

You would if that RGB value were in the content to begin with. And we have every reason to believe with near certainty that such values exist. The discussion about whether or not they exist in content is not an interesting discussion, it's fairly clear-cut: they exist. The interesting discussion is the relevance of that data, the impact of clipping it, and how one might choose to align their display's white point on a digital display. That's an intriguing discussion to have, but we can't get to that point if you're stuck in a world where completely reasonable video codewords are too much for you to accept even in the face of much evidence.

Quote:


So... you should be able to encode RGB 100,100,100 to whatever the YCbCr equivalent is then decode that YCbCr triplet and get 100, 100, 100 again. If every RGB value you encode (and there is NO reason the data should EVER be outside that range) is between 16,16,16 and 235,235,235 you should never, ever, create a YCbCr triplet that decodes to any out of range R, G, or B coordinate.

Well, you're right that if you clip RGB to 16-235 during mastering, you won't really get much outside that during playback if you don't do any processing to the video. But that's not reality. The content isn't clipped, because there is no reason at all to clip it. The examples Stacey posted illustrate just how not-clipped the content is, in this case on BD.

Quote:


[I want to re-iterate a concept you seem to have difficulty with - maybe I can make it simple enough... you claim when you apply a digital tool to data that the tool can/will somehow create out of range digital data.

You're changing the subject again. Lets leave behind digital "magic-math" or whatever it is you want to call them (since you have a problem accepting the existence of digital filters) since that's a couple steps beyond where we are right now which is merely the presence of actual image data, not post-processed samples which may occur. Shall we stick to just what's in the content, and not what happens when we process it later?

Quote:


Let's use an example of 16-bit RGB color data being our original... either from a digital cinema camera or from scanned film. 0 is black, 65535 is peak white.

Except the content's black won't be right at 0 because if you're just a little bit off, then you're hosed when it all is clipped. When it's mastered then choices are made about where in the content you find the black point and the white point.

Quote:


And our editing station knows we are working in 16-bit space. In this situation, you can apply any digital tool or filter or anything you want to do to an image and you are NEVER going to get data outside the 0-65535 range.

Obviously. But you could produce values outside the black/white range wherever that might be in the 16-bit range as it's being worked on.

Quote:


Because the editing station is in 16-bit space and you just can't produce data outside the that range no matter WHAT the digital editing tool does. Now... you may manipulate the tool in ways that produce results you don't like - NO QUESTION. But the data will still be in the 0-65535 range. Now, let's say you are done editing and it's time to prepare your data for your digital master - DVD or Blu-ray. The workstation will handle the conversion for you... down converting from 16-bits to 8-bits AND changing from "full range" (i.e. 0-65535 or 0-255) to the limited 16-235 range used for consumer video. After this is done, no digital editing tools are used on the 16-235 data so there is no chance for it to be carrying data below 16 or above 235. Could somebody screw something up? Probably. Is it "normal" for "recent" (say since 2000) consumer video to have content below 16 or above 235. If it is "normal" it is only because somebody is going out of their way to get there.

All the movies Stacey just posted have been released since 2000. I don't recall Blu-ray being a format in the year 2000...

Or what about the Cars image posted previously. You earlier were hell-bent convinced that it was only possible to produce these "erroneous" "out of bounds" values with old analog equipment. I wonder where we find any analog film or analog workflow in producing Cars... I was under the impression that Cars was an imaginary world created digitally, but perhaps I was mistaken about that and you've actually seen talking cars in your day-to-day travels?

Quote:


There is absolutely NO REASON for data to exist in modern masters that's below 16 or above 235 - somebody has to be forcing data to exceed the easily controlled limits - editing workstations are fully capable of producing 16-235 data without exceptions. Your presumption that out of range data exists seems predicated on 16-235 data being edited when that's not the case - high-bit data is edited and only converted to 16-235 8-bits once there is no more editing, no more filtering, no more changes at all.]

Sure, but you just have it backwards is all. The lack of reason exists for clipping content to 16-235 when Rec709 explicitly allocates the values outside of that reference range for video data. How do you get from "video data" to "out-of-gamut." That's why it's not clipped in the studio and why it's not clipped on the content we get in the home.

My presumption that such data exists is because I've actually looked at DVD grabs myself and measured them. Or because people like Stacey have posted images like the ones he just posted directly from D5 master tape or from BD that illustrate the same. In other words, my presumption is based on the examination of evidence both sought out by myself, and by evidence provided by others.

Your presumption to the contrary is based on no evidence whatsoever that I can discern. In fact, you explicitly describe your complete lack of exploration of any evidence, which is usually characterizes as a state of 'unknowing' or ignorance:

From: https://www.avsforum.com/avs-vb/showt...4#post16172514

Quote:
Originally Posted by Doug Blackburn View Post

In fact, now that I think about it real hard, I never measured any data off any disc.

And yet you're still here arguing about what is actually on the disc or not on the disc. The disc could be filled with lime jello for all you know, but you don't know until you examine the evidence. And until you are actually willing to do that, that is engage in at least some basic scientific endeavor, you're never going to get anywhere.

Minds are like parachutes, yes it's important how you pack them, but it's a lot more important when they open...
ChrisWiggles is offline  
post #38 of 165 Old 05-20-2009, 02:44 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,735
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 26
Quote:
Originally Posted by kamui View Post

Now I'm confused. You "DON'T" want to clip below 16? I thought that you wanted to perserve whiter than white, but clip blacker than blacks.

In other words, you should be able to see visible differences in everything between 235 and 254, but everything from 1 to 16 should be the same "blackness".

What I usually do is take a black 16IRE pattern in 4:3 format and make it match the pillarbox sides of a 16:9 screen, so that 16IRE matches the blackest blacks my monitor can produce. Is this incorrect?

Sorry that I'm going the opposite way of the thread title :P

-Brian

Obviously I won't speak for don, but I assume that he meant in the signal path, not during display calibration on a digital display. You make 16 basically as black as the display can go while still preserving 17. He's not saying you calibrat black level to level 1, you still calibrate to 16, but there is not any good reason to clip off the content in the playback chain at 16 previous to the display. And there are some somewhat obscure processing reasons why you wouldn't want to either, though admittedly the visual impact of that would be pretty hard to see.
ChrisWiggles is offline  
post #39 of 165 Old 05-20-2009, 02:46 PM
Senior Member
 
kamui's Avatar
 
Join Date: Mar 2005
Location: Fort Myers, FL.
Posts: 312
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 38 Post(s)
Liked: 18
Ahh, that makes sense, I got lazy and didn't read the entire tread, probably would have helped me realize this. Thank you.

-Brian

sspears:
(Rules? Rules? In Television???? Thou shalt not oversaturateth red. RGB shall be thy rod and thy staff; there shall be no other color space before thee.)
kamui is offline  
post #40 of 165 Old 05-20-2009, 02:54 PM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:
Originally Posted by kamui View Post

Now I'm confused. You "DON'T" want to clip below 16? I thought that you wanted to perserve whiter than white, but clip blacker than blacks.

Sorry, I just meant the player/display shouldn't throw away the below-reference values and refuse to display them, even if you turn the brightness up. Certainly it makes calibration somewhat harder, and I'm at least open to the possibility that there could be subtle second-order effects of clipping those values.

But you're right - post-calibration all the values from 1-16 should end up producing the same value on your display and that value should be the blackest value the display can produce. Again, modulo some subtle potential effects when scaling and so forth.

Again, bottom line is simple: make the display act like a BVM. If there's any controversy, just check what a BVM does and copy that. BVM's don't hard-clip the values below 16 - they're there if you turn the brightness up. So just on general principles I'm against hard-clipping in consumer displays. If you turn up the brightness, the below-reference values should show up. And then you should turn the brightness back down.

The crazy thing to me is that someone had to go out of their way to clip outside the reference range in the displays that do so. It doesn't add anything of value. It doesn't make processing easier, or storage smaller, or chips cheaper. It's just a dumb error caused by misunderstanding the specs. It's very frustrating.

Don
dmunsil is offline  
post #41 of 165 Old 05-20-2009, 03:15 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,735
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 26
Quote:
Originally Posted by dlarsen View Post

Legal, schmegal.

His statement and my point are about valid and invalid. The quoted statement made no reference to 'legal' anything. I think valid and invalid are pretty well understood terms and concepts and you're introducing legal here as a red herring.

No, I'm not dave, and that was not at all my intention. In RGB, valid is the same thing as legal we the way he uses those terms. In component, they're not the same thing. There is no attempt to obfuscate anything.

Quote:


One theoretical advantage I can see with sRGB is there is no concept, room, allowances or reservations made for any INVALID and OUT-OF-GAMUT 'data' or errors. (BTB/WTW) It doesn't reserve 36% of it's codeword volume for such INVALID data or errors. sRGB uses all of it's 8 bits to represent VALID, IN-GAMUT data.

So how do you handle specular highlights or other bright colors? Where do you put a diffuse white in an image? You don't put diffuse white at 255. You put it somewhere lower than that so you have some room above that to preserve some semblance of highlight detail. But does sRGB standard tell you to do that? No. It just says that reference white is at 255. So if your diffuse reference white in the scene is lower than 255, you're essentially doing the same thing in practice (by being intelligent) as what Studio video levels describes.

Quote:


Which seems better to you- A 8 bit color space with 16.7 million valid, in-gamut codewords (sRGB) or one with 10.6 million valid, in-gamut codewords (StudioRGB)?

Dave

Reason and thought mainly seems better to me. This isn't a war about which is better per se, but understanding them. sRGB has less dynamic range than Studio video for instance. You never bring that up, because it is designed to handle highlight details and not just brick-ceiling at reference white. To achieve the same thing in sRGB you essentially end up with a histogram that would look just like it would at studio levels in order to make some room above a diffuse white for highlight details. So in the end, they're not really that different in practice if you reasonably understand them.
ChrisWiggles is offline  
post #42 of 165 Old 05-20-2009, 03:21 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,735
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 26
Quote:
Originally Posted by dlarsen View Post

I note that in a few of the images Stacy has posted, there are many BTB pixels outside of the active image area and well into the encoded black bars area. How could these pixels possibly even be a candidate for valid image information? If we agree that the encoded black bars should be black, then all these BTB pixels that are out of the active frame area can ONLY be considered as errors. They should be 16,16,16 and they aren't.

Should they? I think it would be better if they were down as far as you could go, at 1 or 0. That's just my opinion.

[quote]If this was transformed and rendered to sRGB levels, these errors in the encoded black bars would be corrected.[/quote

I'm always intrigued how fixated you are on the video data in black bars. They're black bars. They're not part of the picture. Generally when I watch a movie, I pay attention to what's going on in the image, not the bars. That's just me...

Quote:


Also, if it's argued to be a valid extension of image information, why the need to invert (and destroy) the very data in question? Why not post it otherwise unaltered in A/B manner such that it can be contrasted and compared?

Dave

Like you did? And I immediately pointed out, blindly, which was clipped and which wasn't without even measuring them?

Stacey was providing images that illustrate graphically which values exceed the 16-235 range in RGB.

Be reasonable dave. I know you're a smart guy, but you sure as heck are stubborn. I don't know why studio levels did you wrong, but you sure have had an ax to grind for a long time.

And you still won't provide any explanation as to why you'd want to clip studio levels.
ChrisWiggles is offline  
post #43 of 165 Old 05-20-2009, 03:30 PM
Advanced Member
 
dovercat's Avatar
 
Join Date: Apr 2008
Posts: 739
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 104 Post(s)
Liked: 55
Chris I posted a question back in January in your Go-to Guide for Source Options. I will repeat it hear as this thread is more recent and see what your and others opinions are. As my understanding of blacker than black and whiter than white seems to be flawed if they are desirable in a digital display.

Below is my limited understanding, is it correct?

In the Pal analogue signal there is a blanking interval to give the display time to move to the next line or the next frame of the image. In this blanking time there are sync signals. Also during this blanking time the analogue signal is clamped to 0volts by using a capacitor. Because if it is not clamped the signal will wander and the black level in the image will not remain constant (since white is created by higher voltage, a bright image could drag the bottom of the signal up from 0v). If this clamping extends into the part of the signal that syncs black it is called a black level clamp, PAL black is 0volts so it is being clamped to 0volts.

I do not understand how if the above is correct, why blacker than black should or could be passed through the system. My understanding is it was neccessary in analogue systems because they might not have very good clamping of the signal or it might wander up as it displayed a image line with a high average brightness, so some tolerance to black level wandering was needed. Whiter than white information is also I think present in analogue systems because signal level could wander up, at the begining of the line whiter than white would equal white, but at the end of the line white could equal white, if whiter than white was not present in the signal white would be grey at the begining of the line. So my understanding was blacker than black and whiter than white are holdovers from the analogue age when image brightness might not hold steady across the image, you might need to drag an analogue wave signal down to black or up to white when the voltage of that signal could be wandering about.

In a perfect CRT I think the idea is that it is not needed as the signal level does not wander, unfortunately people did not have perfect CRTs so it was present in the analogue signal, but on a perfect CRT the information would not be displayed, in effect it would be clipping below black information and whiter than white information would be white anyway. Calibrating to incorporate blacker than black and whiter than white apears to me as calibrating to be equivelent to the worst crt display it was designed to cope with. I thought they only existed in digital as a legacy to analogue and served no usefull purpose with digital displays.

If blacker than black is desirable up to the digital non-crt display, and according to some displaying some blacker than black information is desirable. What visible negative impact does it have if the source is clipping it and are sources that clip it failing to meet the consumer spec and of a faulty design for doing this.

If whiter than white if it is desirable to actually display. What negative impact on image quality is there from clipping it and again are sources that clip it failing to meet the consumer spec and of a faulty design.

If I do not display blacker than black or whiter than white I gain contrast, which improves perceived image quality. What do I gain if I display them.
dovercat is offline  
post #44 of 165 Old 05-20-2009, 03:52 PM
AVS Forum Special Member
 
sspears's Avatar
 
Join Date: Feb 1999
Location: Sammamish, WA, USA
Posts: 5,320
Mentioned: 23 Post(s)
Tagged: 0 Thread(s)
Quoted: 315 Post(s)
Liked: 512
Quote:


On a CRT, a really low below-black value has the potential to affect the next pixel, pushing it up just a tad because of filtering, internal ringing, etc.

One point Joe makes in his tutorial, on DVE, is that a CRT, even a BVM does not have perfect linearity from side-to-side. You might set black level using a specific APL pattern, but as the APL goes up or down, black floats. This is one reason Joe actually sets black level one click above where one might set it on the Samsung.

It is worth noting that Fotokem has the Samsung in house for use. While the Panasonic 11 series pro plasma has been replacing BVMs all throughout Hollywood, so has the HP Dream Color LCD. A friend just calibrated a bunch of the HPs for a studio. He used the BVM in the room to model the HP and matched them up. The black level was poor on the HP, compared to the BVM, but color gamut and gamma were spot on. A 2.4 gamma I might add. These displays were all calibrated to ensure you see RGB values up through 254 in 8-bit space.

Stacey Spears
Co-Creator, Spears & Munsil UHD HDR Benchmark
sspears is offline  
post #45 of 165 Old 05-20-2009, 04:13 PM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by ChrisWiggles View Post

No, I'm not dave, and that was not at all my intention. In RGB, valid is the same thing as legal we the way he uses those terms. In component, they're not the same thing. There is no attempt to obfuscate anything.

Huh? Again, there was no mention of legal, only valid. Legal and valid are not the same thing. Read it again.
Quote:
Originally Posted by from Poynton View Post

The so-called valid colors encompass the volume that is spanned when each R’G’B’ component ranges from reference black to reference white. In Rec. 601, each component has 219 steps (risers) – that is, 220 levels. That gives 220*220*×220, or 10648000 colors: About 64% of the total volume of codewords is valid.

Each and EVERY BTB/WTW triad is SPECIFICALLY EXCLUDED from his classification of being valid. If something is excluded as being valid, then it is INVALID!

Quote:
Originally Posted by ChrisWiggles View Post

So how do you handle specular highlights or other bright colors? Where do you put a diffuse white in an image? You don't put diffuse white at 255. You put it somewhere lower than that so you have some room above that to preserve some semblance of highlight detail. But does sRGB standard tell you to do that? No. It just says that reference white is at 255. So if your diffuse reference white in the scene is lower than 255, you're essentially doing the same thing in practice (by being intelligent) as what Studio video levels describes.

Well, new video standards are also totally eliminating the notion of needing head/toe room for quality digital imaging. If head and toe room are sooo precious and necessary, why are new standards like xvYCC totally eliminating it?

Quote:
Originally Posted by Wikipedia View Post

“xvYCC uses the full range of values (0 to 255 in an 8-bit space) to represent colors. In BT.601 and BT.709, RGB colors are represented only by 8-bit values from 16 to 235. This limited range was established to allow for undershoot and overshoot which were attributes of analog TV signaling. With digital TV signaling, there is no undershoot or overshoot, and the values from 0-15 and 236-255 can be used to represent colors.”

Quote:
Originally Posted by ChrisWiggles View Post

I'm always intrigued how fixated you are on the video data in black bars. They're black bars. They're not part of the picture.

Exactly my point. Black bars that have BTB pixels in them are ERRORS and can’t possibly be valid information. Clearly to any rational person, a BTB pixel that is in the encoded black bars can’t be anything other than an error but they are often put forth as an example of valid data. BTB errors in the encoded black bars can only exist when transforming and rendering to StudioRGB levels. Such errors can’t exist with sRGB.

Quote:
Originally Posted by ChrisWiggles View Post

Like you did? And I immediately pointed out, blindly, which was clipped and which wasn't without even measuring them?

Yes, I found it odd (not really) that you were the ONLY one who could perceive any difference. Gifted or… Everyone else who reported reported perceiving absolutely zero difference. Even Darrin. I have several more A/B images posted and no one has reported perceiving any differences at all. I can do it again with the images posted here if they get posted up unaltered and with the very pixels in question not destroyed and drastically altered. Stacy and Don could also post such images in a side-by-side, lossless, full-frame A/B manner so they could be contrasted and compared side-by-side, lossless, and full-frame. If they wanted to. I can only guess as to the reason they don't post them in such an otherwise unaltered A/B manner.

Dave
dlarsen is offline  
post #46 of 165 Old 05-20-2009, 07:38 PM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:


If head and toe room are sooo precious and necessary, why are new standards like xvYCC totally eliminating it?

xvYCC doesn't change the interpretation of headroom; it's essentially the same. Only toeroom is changed. Wikipedia is wrong on that subject. The gamut expansion is entirely from the negative values.

And xvYCC retains the same reference white, the same gamma curve, same everything for every code value above 16. They just acknowledge that the range between reference and peak is available for use, which IMO was always the case, especially for shiny discs. Standard practice for use of the upper range isn't going to change much, except that some mastering houses will turn off their limiter. Which they could have done anyway if the content wasn't intended for broadcast.

Look, part of your argument appears to be that nowadays we don't need headroom and toeroom and we could just use the PC code range and let everyone set their "reference white" however they see fit. I think on a technical level that's basically true, though in practice I think it would lead to video tending to look worse (because people would handle peak clipping badly). So what? We have what we have. There's no practical way to switch over at this point, and very little visually to gain. In practice, you'd really only get back the below-reference code values, and that's not going to add much. A teensy bit more resistance to banding. Big whoop.

If we didn't have the extra code values we wouldn't be able to produce a wide-gamut format that's backward compatible, and I do think once wide-gamut displays and content comes out that's going to be a really nice improvement. People love saturation. Well, everyone except Stacey.

Don
dmunsil is offline  
post #47 of 165 Old 05-20-2009, 10:31 PM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by dmunsil View Post

Some of the above-reference material is basically unimportant. It just doesn't change the image in any visible way.

I’d agree with this. This has been my perception in every lossless, full-frame, otherwise unaltered A/B side-by-side comparison I’ve seen or posed and I’ve posted dozens of them over the years.
Quote:
Originally Posted by dmunsil View Post

Some of it, on the other hand, is easily visible when you compare clipped to not-clipped.

I’ve yet to see this posted or demonstrated in a full-frame, lossless, non-GCI, A/B side-by-side. Surely since you and Stacey make this claim, you have done these full-frame, lossless, side-by-side A/B comparisons? Why don’t you (or Stacey) post two full-frame, lossless, non-GCI, A/B side-by-side images that demonstrates how this is ‘easily visible’ when one compares clipped to not-clipped with normal mainstream material from a mainstream studio shiny disk?

It seems to me that if this was as ‘easily visible’ as you claim, then there would be hundreds of such side-by-side A/B images posted here and there wouldn’t be ANY debate. No one would even bother posting caps at sRGB levels because everyone would know how inferior they would appear. Yet, most all of the hundreds of excellent caps posted here by Xylon and others ARE capped and posted at sRGB levels and contain ZERO BTB/WTW. If anything perceptible was there, then it's been clamped out.

Yes, Chris claimed that the caps I did and posted from the DVD “Ali” demonstrated it- but he was the alone in this and all the other responders including myself and Darrin could perceive absolutely zero perceptible difference in the half a dozen or more A/B images I posted. Unfortunately, that thread got deleted but Chris is certainly welcome to post them up anytime he chooses to do so. He could rent the DVD Ali, cap and post them up himself just like I did but to date he hasn’t done so.

Why aren’t the images Stacey posted here done A/B side-by-side? Why the need to invert and obliterate the very data (or non-data) in question? Could it be because that would only demonstrate your first observation that I quoted, rather than your second claim that I quoted?

Dave
dlarsen is offline  
post #48 of 165 Old 05-20-2009, 11:18 PM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
One other note: The altered images that Stacey has posted inverts (and obliterates) when any ONE component is over 235 or under 16. If this range were really being used deliberately and intentionally as an extension of ‘valid information’ I’d think these would be much larger than a tick or two away from 16/235 and would involve more than just one component.

I think you’ll find it interesting and VERY rare where you’ll find instances where ALL THREE components are <16 or >235. Try highlighting instances where all three components (R,G,B) are <16 or >235 in mainstream material. I think the fact that it is VERY rare that all three components excurse beyond the reference bounds, speaks volumes about their true nature.

Dave
dlarsen is offline  
post #49 of 165 Old 05-20-2009, 11:35 PM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:
Originally Posted by dlarsen View Post

Surely since you and Stacey make this claim, you have done these full-frame, lossless, side-by-side A/B comparisons?

Yes. We've had this argument before, going all the way back to 2004. We posted examples then, and you pretty much ignored them. After you and JeffY kept ignoring the images we posted, Stacey and I stopped arguing with you. Chris apparently has more stamina.

I'm sorry, but if the post I made back in 2004 of side-by-side zoom of a frame from Toy Story clearly showing that the image is visibly changed when you clip in RGB didn't convince you, then I'm done. You may find this difficult to believe, but we didn't go hog wild looking at tons of films to find that example. It was quite literally the first frame we looked at in the first film we looked at. If you feel like it, you can do the same work we did. We looked at enough movies to convince us that every movie has above-reference values in RGB, and then we quit, because (and again, this may surprise you), we did it to satisfy ourselves, not to convince people on the internet, especially people who apparently think that if you use a Pixar movie to make a point, it doesn't count.

Stacey posted a frame from Cars in this thread. It has a bunch of above-reference pixels. I know for a fact that he didn't spend any time finding that frame; he just grabbed a frame he happened to be looking at because he was working on tracking down a banding issue. If you feel like it, you can take that frame and clip it and switch back and forth between clipped and non-clipped. I haven't done it, but my guess is that there will be subtle color shifts between the two frames. Does it ruin the presentation? No, not really. But what does that prove? You can do a whole bunch of things to the image and not ruin the movie. It's still a good movie even if you watch it with the color decoder miscalibrated, or the brightness two clicks too low, or any number of small issues. But the point of being an enthusiast is to take the time to get everything right.

So if your point is that you can clip above 235 and the movies are still enjoyable, then we have no disagreement. If you are arguing that it doesn't change the image, then you are wrong. We have already demonstrated it via actual side-by-side images. But they were CGI images, so they don't count? That's seriously your argument? Oy.
dmunsil is offline  
post #50 of 165 Old 05-21-2009, 12:11 AM
AVS Forum Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,912
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Yes, CGI can be a special case (Poynton make references to it's special case work flow and it's use of SuperBlack) and ToyStory I was a special, special case. It's was shown to have excursions up to and including 255 and even you and Stacy agreed that 255 was reserved and shouldn't be used.

FWIW, here are some full-frame, non-cgi, side-by-side A/B comparisons.

https://www.avsforum.com/avs-vb/showt...5#post16153195

https://www.avsforum.com/avs-vb/showt...2#post16023752

And many, many outstanding images at sRGB levels (with zero BTB/WTW possible) and without any purported banding issues due to sRGB rendering.

https://www.avsforum.com/avs-vb/showt...0#post15970830





Dave
dlarsen is offline  
post #51 of 165 Old 05-21-2009, 02:16 AM
ADU
AVS Forum Special Member
 
ADU's Avatar
 
Join Date: Apr 2003
Posts: 9,342
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 687 Post(s)
Liked: 1423
Quote:
Originally Posted by ChrisWiggles View Post




The histograms in these images do not look right to me.

The images are so dark that the majority of the RGB Color and Luma information should be in the darker range, rather than more or less evenly distributed the way these graphs show. I think you've either got them mixed up with another image's data, or your software appears to be grossly distorting the information in some way.

ADU
ADU is offline  
post #52 of 165 Old 05-21-2009, 03:46 AM
Advanced Member
 
dovercat's Avatar
 
Join Date: Apr 2008
Posts: 739
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 104 Post(s)
Liked: 55
So if I have got this right, those in favour of displaying whiter than white are saying not doing so causes subtle color shifts and possibly slightly more banding. Obviously it also alters average image brightness and contrast which can both effect perceived image qualtiy as well. While the few in favour of blacker than black making it to the display but not being displayed, are saying it should not be displayed but its presence in the signal path has an effect on how the displays rescaling algorithims calculate the scaled up image.

I will check out whiter than white myself looking at a greyscale ramp test screen, then comparing several movie scenes from several different films I am familiar with. To see which I prefer.

Any recommendations as to films or scenes where the differences are most blantantly obvious.

Out of intrest why does everyone quote Poynton, and seemingly only Ponyton. Has everyone read his book as required or recommended reading on some college course or is he recommended reading in ISF or THX courses, or does he just have the most web presence? What about the AV engineering books that cover display technologies and the theories behind them. Why are Ponytons opinions taken as definitive, when other authors do not necessarily share them.
dovercat is offline  
post #53 of 165 Old 05-21-2009, 08:19 AM
 
tbrunet's Avatar
 
Join Date: Jul 2005
Location: Savannah, GA
Posts: 3,257
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
http://www.fsl.cs.sunysb.edu/~sean/misc/video.pdf

Charles Poynton is an independent contractor specializing in the physics, mathematics, and engineering of digital color imaging systems, including digital video, HDTV, and digital cinema (D-cinema). Charles was a key contributor to current digital video and HDTV studio standards.

In the early 1980s, Charles designed and built the digital video equipment used by NASA to convert video from the Space Shuttle into NTSC. In 1990, he initiated Sun Microsystems' HDTV research project, and introduced color management technology to Sun. He was responsible for Sun's founding membership in what later became the International Color Consortium (ICC).
tbrunet is offline  
post #54 of 165 Old 05-21-2009, 08:55 AM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:
Originally Posted by dlarsen View Post

Yes, CGI can be a special case (Poynton make references to it's special case work flow and it's use of SuperBlack) and ToyStory I was a special, special case. It's was shown to have excursions up to and including 255 and even you and Stacy agreed that 255 was reserved and shouldn't be used.

My experience says that Toy Story isn't particularly unusual.

Quote:

Great. You've demonstrated that on those particular frames, viewed side-by-side, it's pretty much impossible to tell whether or not the image was clipped. But side-by-side full frame is not the most revealing. Nor are specular highlights the places that necessarily are the most visible.

The best way to see differences is to switch back and forth between the two images in the same space, which is what I did with Toy Story back in 2004. Then I cut out a chunk that demonstrated my point, zoomed in on it, and posted it. You continue to assert that it's a fluke or something, which is certainly your right.

However, here's the history of this argument, from my perspective:

- We point out, correctly, that above-reference information is not clipped by professional video monitors.
- You and a handful of other people assert that there is no above-reference information on DVD
- We demonstrate that there is above-reference information on essentially all DVDs
- You say that it's just noise, it's not visible.
- We show a clear example of something that is not overshoots, not noise, and is visible to anyone who cares to look at it.
- You say it doesn't count because it's CGI

I'm sorry, but at this point the goalposts have been moved too many times. I see from the threads you linked that tbrunet still says that no movie really contains above-reference data, something that is simply not true and has been demonstrated time and time again, by multiple people.

It appears to me that you are not actually arguing in good faith. You seem to have a chip on your shoulder about this issue. If I find you a non-CGI example, I expect that you'll then argue that it's not relevant because of some new objection, because that's been the history of this argument. You say we're wrong, I demonstrate that we're not, and you move the goalposts. Lather, rinse, repeat. And frankly, I have work to do. It takes time to generate these demonstrations, and when they seem to have no effect and get dismissed with random objections like "it's CGI," I don't feel any obligation to take time out to find you more examples.

I'm happy to meet you halfway and say that people may have used rhetoric that was too strong to describe what happens when you clip the high range. It's an aesthetic judgement. I certainly don't want to see something different from what the mastering engineers saw; I want to see the same thing. Once I established to my satisfaction that there were good movies that contained above-reference picture information (not just noise and overshoots), then I felt like that was enough for me.

Quote:


And many, many outstanding images at sRGB levels (with zero BTB/WTW possible) and without any purported banding issues due to sRGB rendering.

https://www.avsforum.com/avs-vb/showt...0#post15970830

I don't see any banding in those images, and wouldn't expect to. The grain works like dither. And you're right that most movies have enough grain most of the time to hide any banding. But they don't always. If you want a better test, look at the dreaded CGI movies, and movies shot digitally. Movies that have had DNR applied will be more susceptible to banding as well. It's not complicated; in order for banding to show, you need a smooth clean gradient. It seems to happen most often in skies and smoke, but it's not 100%; lots of movies have enough noise in the sky or smoke that you can probably knock out every other code value and they'll still look fine. Heck, Stacey and I demonstrated that with proper dithering you could take video down to 2 bits per channel (well, 5 codes, so 2.1 bits) without banding.

If someone has said or implied that banding will *always* result from range expansion, then I think they should walk that back, because it's not true. However, it will sometimes result from range expansion. This is easily demonstrated from test patterns, but certainly I've seen it happen with real picture material. I see it, for example, when I'm editing photos I've taken. Most of them are OK when you do picture edits that produce "broken comb" histograms, but some pictures show banding clearly when you expand the range. And then you have to do some extra work to fix the banding. Maybe this never happens to you, and that's great.

You seem to think that I owe you proof that it happens, but I don't think I do. Frankly if you're happy with the picture you're getting, and are not seeing banding or clipping, I'm happy that you're happy.

And your basic point, as I see it, that most movies will look fine clipped and expanded, is essentially true. But again, this would be true as well for all kinds of things you could do. You can watch a movie with a factory-set TV right off the Best Buy floor with oversaturated colors and mis-set brightness, and most people will be fine with it.

And hey, over the years I've mellowed on these issues. It's not a crime to watch movies the way you like to watch them. If you like the extra contrast you get from expanding the range, then more power to you. I feel obliged to mention that it's not to spec and you may see subtle color shifts in saturated colors and loss of detail on bright white objects, or banding in some material and you're free to say that you haven't seen them and don't expect to. To each his own.
dmunsil is offline  
post #55 of 165 Old 05-21-2009, 09:02 AM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:
Originally Posted by ADU View Post

The histograms in these images do not look right to me.

The vertical range is log. It's more useful than a linear range when you have such huge disparities in absolute numbers of pixels at each code level. It's fairly common in HDR imaging software, and we like using it in our tools.
dmunsil is offline  
post #56 of 165 Old 05-21-2009, 09:17 AM
Member
 
billcushman's Avatar
 
Join Date: Mar 2009
Location: Houston TX
Posts: 46
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
It is sad that manufacturers such as JVC continue to get this wrong. When you select Expand on JVC projectors, the peak white (254) is properly reproduced, but the Brightness and Contrast controls must be readjusted from their default values. Because this is beyond the skills of non-technical users without a test disc, these users are probably better off leaving everthing including the Normal/Expand setting at the default values, rather than incorrectly setting the user controls, thus they have clipped whites.

Don Munsil and Stacey Spears have stated the standards and shown examples very clearly. They have been very patient in trying to explain this issue to those who misunderstand the standards for proper reproduction of digital video.
billcushman is offline  
post #57 of 165 Old 05-21-2009, 09:22 AM
AVS Forum Special Member
 
dmunsil's Avatar
 
Join Date: Mar 2000
Location: Kirkland, WA, USA
Posts: 1,005
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 30
Quote:
Originally Posted by dovercat View Post

So if I have got this right, those in favour of displaying whiter than white are saying not doing so causes subtle color shifts and possibly slightly more banding.

Yes. And that pro monitors don't clip above-reference.

Quote:


Obviously it also alters average image brightness and contrast which can both effect perceived image qualtiy as well.

Well, absolute display brightness is not usually considered a picture quality issue. If you can get reference white to a reasonable brightness, say 16 fL, in a dark room, you're in good shape. Modern displays usually have no problem producing that kind of brightness while not clipping the above-reference values.

Quote:


While the few in favour of blacker than black making it to the display but not being displayed, are saying it should not be displayed but its presence in the signal path has an effect on how the displays rescaling algorithims calculate the scaled up image.

I'm certainly in favor of below-black making it to the display, just because it's the way the system is supposed to work. But clipping it is significantly less of a problem than clipping above-reference.

Quote:


Out of intrest why does everyone quote Poynton, and seemingly only Ponyton. Has everyone read his book as required or recommended reading on some college course or is he recommended reading in ISF or THX courses, or does he just have the most web presence? What about the AV engineering books that cover display technologies and the theories behind them. Why are Ponytons opinions taken as definitive, when other authors do not necessarily share them.

Poynton's books are the best, though Keith Jack also has a good book on video standards and there are others. Poynton's strength is color theory, but he's good on all aspects of video.

Dave is absolutely correct that Poynton feels strongly that above-reference should not contain picture information. However, Charles feels strongly about lots of things that aren't standard practice. You should have seen the knock-down drag-out fight on the Hollywood Post Alliance forum about whether it was better to refer to 1080i as 1080i30 or 1080i60. Makes this argument seem like a pleasant discussion over tea. Suffice it to say that even though I love Poynton's books, have learned immense amounts from him, his articles, his books, and so forth, I don't agree with him on everything.

Whatever Poynton believes, it's abundantly clear that real world video does have visible picture information in the above-reference area. On some films, it's barely there; just overshoots and noise. (Even then, I think overshoots and noise can be visible.) On other films, there's real honest-to-God picture info. You can clip it off and it's not necessarily going to look terrible. Some clouds will get pasty-looking from time to time. Some colors won't look quite so saturated. It's visible, but is it important? That's kind of an aesthetic judgement.

But the bottom line for me is this: the mastering engineers saw it. The picture they worked on and approved had visible above-reference values. I want the same picture they have, not a close approximation. That's the goal.

If your goal is to get the best-looking picture to your eye, which I think lots of people do, then adherence to standards may be less important, and there's no law against that.
dmunsil is offline  
post #58 of 165 Old 05-21-2009, 09:43 AM
 
tbrunet's Avatar
 
Join Date: Jul 2005
Location: Savannah, GA
Posts: 3,257
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by dmunsil View Post

Whatever Poynton believes, it's abundantly clear that real world video does have visible picture information in the above-reference area. On some films, it's barely there; just overshoots and noise. (Even then, I think overshoots and noise can be visible.)

Overshoot and filter ring (analog or digital domain) & noise are one in the same. Color space gamut errors. One cannot reverse engineer xvYCC (or expand color gamut) with out encoding them for that specific purpose.
Quote:
Originally Posted by dmunsil View Post

But the bottom line for me is this: the mastering engineers saw it.

According to SMPTE Conference VP John Luff what is seen as a desired effect (WTW/BTB) on the master eng CRT monitor will more than likely not be reproduced on other display technologies or even other CRT’s for that mater. John Luff also commented and that it was explained to him that the ITU-R 601 transform itself has a built in filter if you will ..to either mask or dampen high frequency filter overshoot and ringing.

This debate predated HD / ITU-R BT.709
tbrunet is offline  
post #59 of 165 Old 05-21-2009, 09:58 AM
ADU
AVS Forum Special Member
 
ADU's Avatar
 
Join Date: Apr 2003
Posts: 9,342
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 687 Post(s)
Liked: 1423
Quote:
Originally Posted by dmunsil View Post

The vertical range is log. It's more useful than a linear range when you have such huge disparities in absolute numbers of pixels at each code level. It's fairly common in HDR imaging software, and we like using it in our tools.

I see. Thanks for the reply Don.

I'm afraid that a logged graph may be giving folks an incorrect impression of how much information there is in the WTW/BTB regions. Do you have linear histograms so we can get a better idea just how much of the picture info falls outside the 16-235 range on these images?

Quote:






ADU
ADU is offline  
post #60 of 165 Old 05-21-2009, 10:14 AM
 
Doug Blackburn's Avatar
 
Join Date: May 2008
Location: San Francisco - East Bay area
Posts: 3,587
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 188 Post(s)
Liked: 354
Quote:
Originally Posted by dmunsil View Post

Oh, and whoever said that above-reference values are out of gamut was wrong; increasing the brightness range of R', G', and B' doesn't move them out of gamut. {235, 16, 16} is the same color as {254, 16, 16}, and is way in the far corner of the gamut triangle. One is just a brighter version of the other. Conversely, there is a visible difference between {235, 235, 235} and {254, 235, 235}, but both colors are fully in gamut. One of the basic color principles is that the color of any all-positive combination of three primary colors lies inside the triangle formed by those primaries.

Don

Wait a minute... If you are working with the consumer video format and your range of RGB values is supposed to be "16-235" (let's ignore for a moment that higher or lower values are "permitted") -- any coordinate containing a number higher than 235 or lower than 16 might be called "out of gamut" for the purposes of this discussion even though the spec is not hard limited at 16 and 235. I think the issue here really is... "What is your gamut?" If you are talking about a gamut defined by 16,16,16 and 235,235,235 (and 16,235, 235 and 16,16,235 etc. etc.) then, yes, 254,235,235 is out of gamut. Yes? The issue is that while 16-235 defines the nominal gamut, there's really nothing technically "wrong", spec-wise, with producing video content with values in the 236-255 or 0-15 ranges - the spec permits it - the spec permits the gamut to be defined by 0,0,0 and 255,255,255 (plus 0,255,255 and 0,0,255 etc.) even though the nominal range is 16-235.

In digital video, data could easily be constrained to 16-235, it just isn't done because it's not "required" by the spec -- even though some display manufacturers are taking the 16-235 range literally (on one or both ends).

The end result is that video content released into a nominally 16-235 format may actually have "real" black at 10 or 6 or 13 and "real" white at 241 or 252. We setup video displays so 16 is black and 235 is white and let every program or disc set arbitrary minimum and maximum digital values because the spec allows that. Essentially, there is no standard except they can't go lower than 0 or higher than 255. (yes, I understand that data below 16 may be more sloppy than important, but it's there, apparently and you won't see it if your black level is 16).

If every video display, even the ones on edititing stations, are set so 16 is black and 235 is white... and we are working with digital video where it would be very easy to constrain the final product to 16-235... we are left with this nebulous world where one disc has peak white at 37 fL (on your display at home) while the next disc has a peak white level at 32 fL. It would be easy for digital editing and mastering to honor the 16-235 range. Discs would meet a well-defined standard so when we adjust the display for black at 16 and white at 235, that's what we always get. I just doesn't happen because the spec doesn't require it even though it would be very easily achieved in digital video.

The frustration is that when there are no hard limits on the working range (like 16-235), we can't have the precision setup we might like to have so we see all the available info without the peak value for white changing from disc to disc to disc.
Doug Blackburn is offline  
Sponsored Links
Advertisement
 
Reply Display Calibration

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