or Connect
AVS › AVS Forum › Display Devices › Digital Hi-End Projectors - $3,000+ USD MSRP › Some people are still being told that 200:1 is all we can see....
New Posts  All Forums:Forum Nav:

Some people are still being told that 200:1 is all we can see.... - Page 8

post #211 of 505
Quote:
Originally Posted by armadillo View Post

Darin: Ha!!! I just did a cool experiment. See for yourself. Make an image in PS or your favorite graphics app. Make it 600x600 with black background. Then create four equally size squares 0f 250x250 and place them on the image so that you reatin a 20 pixel distance. Set the brightness of the squares as topleft: 2% (adjust this value so that you just barely see it against the black background); topright 20%; bottomleft 40%; bottomright 60%. Finally create a fifth square of 250x250 and place right in the middle of the four and make that 100% brightness. You should place this square into a separate frame (or cut to the clipboard).

Now look at the center of the four squares. You should be able to see all of them at the same time, although the top left one will be ever so faint. You can even concentrate on the right or bottom bounds of that faint square to actually see the contrast. Now make the white center square visible (either by activating the layer or pasting it) and you will see that the top left square is gone. Make the white square invisible and the square will slowly reappear.

Edit: You can even even make all four squares faint (I can go up to brightness of 3% on my monitor). As soon as you make the white square visible, they all disappear for me. Even if I try to actively see the boundaries of the squares, which are still there, I cannot see them while the white square is present.

I just did this in paint. The spread between squares isn't 20 pixels, that doesn't seem particularly important, it's roughly 40 pixels. The lowest square at 2% I cannot see even on an all black frame on an LCD monitor. I can on my CRT. So I made the dimmest square about 4% (level 10 in graphics range). I can see all four squares without any difficulty with the presence of the 100% square at the center. All squares are equal size, 250x250 on a black background. On a poor-ANSI display the presence of the white square may wash out and obscure the visibility of the other squares. But on an LCD which has pretty high ANSI, any person with reasonable visual acuity should have no difficulty at all seeing these squares at the same time.

I've attached the crude pattern I whipped up. It compressed down to jpeg to be attached, but that's what I'm looking at. (Note: attached image is for graphics range, NOT video range)


Also, I shouldn't have to note that this test is FAR less relevant than the hand shadow puppet tests described earlier because this test relies so heavily on ANSI CR performance of the display, making it basically irrelevant except as a measure of ANSI CR performance of a display. That we cannot see the dimmest square on many displays has little to do with our vision, and everything to do with poor CR performance of the display.
LL
post #212 of 505
My take on this is the need to realize it is the instantaneous ability of the human eye that is of concern. I have been using a CRT in the living room for years (9500LC, 110 screen). While watching movies, when the screen goes black, I can barely navigate the room, can't see anything. Later at night when I get up (darn medication), I can nearly read and the screen is gray, I can really see it. I really need to get the Theater built. The CRT is not very bright, but things like a bright flash, like lightening in a really dark scene can hurt a little. On the other side, where this topic resides, going instantly from a predominantly high APL to a very low APL, I don't see all for a while (a few seconds or so).

As for the computer screen testing, one thing to try is to generate a stair pattern, black to gray in small steps, then put up a white screen for 15-20-seconds then close the white screen and see how much of the stair pattern you can see. You can then take a measurement of the light output of white and that of the first black line you could or couldn't see and measure calculate the CR. This is a moving target, try it at 5-seconds and at 30-seconds and see the differences.

I am also sure all of us react differently and age can make a difference too.
post #213 of 505
Quote:
Originally Posted by armadillo View Post

That's not a fair situation, since you are separating the contrast and looking at a subimage, not the entire image within your fovea.

You mean someone is using their eyes normally? That's not fair?

Gee whiz! Heavens ta betsy! Lets all bolt our heads to a wooden board and glue our eyes so we can't move them and tape open our eyelids, and watch images the size of a postage stamp so we can only discern 4 pixels, that way we only need a display that can achieve 100:1 between those four pixels! What are we all waiting for? We've been CHEATING this whole time using our whole visual capabilities!

Pretty soon we'll all be pouring acid in our eyes to make sure we're not being unfair in our viewing habits... In the meantime, we should stop making movies in color. We should only make a very small grouping of pixels in the center of the screen in color, the rest should just be B&W since we can't see color very well outside the small foveal area. And since we're all strapped down and glued tight and not moving our eyes around like normal people, we won't be cheating. That's the real movie experience, you all are just too dense to recognize your own vision limits because you're so busy moving your eyes around! And to eliminate the unfair ability of temporal adaptation, we should ensure that there is an absolute minimum of temporal change in the images we view. Preferably there should just be one still image we observe to ensure we're never adjusting temporally, because that's not fair.

Goodness, someone says I can't jump more than an inch off the ground. I should cut my legs off to make sure that's true so I'm not being unfair! Heck, I'll bet I could jump that high just with my arms, I better cut those off two.
post #214 of 505
Hey guys...

So, I've cut off my legs and one arm, but I'm having some difficulty cutting off the last one. Maybe someone can help be out? My neck just isn't that flexible to gnaw it down at the shoulder...

I gnawed it off at the elbow, but I still think I could fling my torso up about an inch if I really tried, and I just wanted to be sure I wasn't being unfair, so I figured I'd go all the way to the shoulder to be absolutely sure. Also, my eyes hurt now because they aren't being lubricated by my eyelids. I'm having trouble seeing much of anything at all. But I'm so happy to be viewing fully within the most extreme limitations of the HVS. It feels so good to be liberated into a world where I can only see 100:1 directly in front of me.

I typed this with my nose.

I'm going to have to figure out a way to immobilize my neck too, because I can still see very large angles by moving my head around, despite having glued my eyeballs fixed in their sockets. I'm not sure how everyone else does that.
post #215 of 505
Chris: I looked at your jpg and that doesn't work on my monitor either. It's not surprising, since I mentioned that I can go to 3% brightness to get the effect, but it fails at 4%. What may be happening is that you need a fairly good LCD monitor to observe this phenomenon. I sit in front of a 30" Apple Cinema Display with a purported CR of 700:1. I am probably not getting the full ratio, since I have brightness and contrast reduced through the monitor's controls. So I interpret you inability to see the effect as meaning that your monitors are probably not that great in the CR department, so you can't effectively shift your HVS system out of the contrast ratio range required to make the square disappear. It would be interesting to know whether the contrast test in the link HHS posted works for you.
post #216 of 505
Quote:
Originally Posted by Gary Lightfoot View Post

I was recently fortunate enough to be invited along to see some high end projectors, and in the conversation I found that at an ISF event/course the host was told that 200:1 contrast is all that we can see. A well known name was also mentioned as the source so I was very surprised that this information is still being thrown around even by people considered to be very knowledgeable in this particular field.

I'm still trying to learn about the subject, and thanks to Darin, Chris, Erik etc, It's starting to sink in (slowly), but although the best example I could give was that we can easily see the difference in contrast capabilities of projectors with just 500:1 and say 2500:1 (LCD vs DLP), it wasn't accepted. I also asked why the projectors had higher on/off than 200:1 if that's all we can see...

Dynamic range was said to be important in the conversation and that was why these particular projectors were so good, it wasn't explained what 'dynamic range' was, so I asked if it was on/off contrast. It was agreed that it was so I suggested that if 200:1 was all we need, why was the dynamic range so important (unless what they actually meant was image brightness and I think it was in the most part)? Both statements seem contradictory to me if that's the case.

I did try to find the context of the statement to see if it was on/off or ANSI contrast, but it did appear that it was indeed on/off contrast.

I often see 100:1 bandied around which I think actually refers to the eye's CSF, but 200:1 may be the 1:200 that Darin mentions here:

http://www.avsforum.com/avs-vb/showt...&&#post7930256

Has anyone else here been told that 200:1 is all that we can see at an ISF course or event, and what examples can be given that are good enough to put doubt into peoples minds regarding that particular number?

Gary

I just discovered this interesting thread and haven't gone through it yet, but have something to contribute (if it hasn't been mentioned before).

In Digital Video and HDTV: Algorithms and Interfaces Charles Poynton says the eye can process a 1000:1 contrast ratio.

However, just 100:1 is sufficient for a TV.

Why? Mainly because nonlinear coding of TV signals (presumably including HDTV signals) reduces the specular highlights in video luminance by a factor of 10:1.

Quote:


At the reproduction device, we can seldom afford to reproduce diffuse white at merely 10% of the maximum luminance of the display, solely to exactly reproduce the luminance levels of the highlights! ... A convincing image can be formed with highlight luminance greatly reduced from its true value. To make effective use of luminance ranges that are typically available in image display systems, highlights must be compressed. (p. 83)

Poynton equates nonlinear coding of the input to a video camera to "tone scale alteration"/"gamma correction." One reason for doing it is to limit specular highlights. Another is to constrain contrast ratio in view of the fact that ordinary TVs have a simultaneous contrast ratio that "is typically less than 100:1 owing to spill light (stray light) in the ambient environment or flare in the display system" (p. 197).

Admittedly, I don't fully understand the reasoning here. The TV's "gamma" function, whose exponent is nominally, say, 2.5, is supposed to (more than) fully offset the "gamma correction exponent" (say, 1/2.2) in the video camera. I don't quite see why that doesn't (more than) restore the original 1000:1 contrast ratio in the scene!
post #217 of 505
Quote:
Originally Posted by epstewart View Post

I just discovered this interesting thread and haven't gone through it yet, but have something to contribute (if it hasn't been mentioned before).

In Digital Video and HDTV: Algorithms and Interfaces Charles Poynton says the eye can process a 1000:1 contrast ratio.

However, just 100:1 is sufficient for a TV.

Why? Mainly because nonlinear coding of TV signals (presumably including HDTV signals) reduces the specular highlights in video luminance by a factor of 10:1.



Poynton equates nonlinear coding of the input to a video camera to "tone scale alteration"/"gamma correction." One reason for doing it is to limit specular highlights. Another is to constrain contrast ratio in view of the fact that ordinary TVs have a simultaneous contrast ratio that "is typically less than 100:1 owing to spill light (stray light) in the ambient environment or flare in the display system" (p. 197).

Admittedly, I don't fully understand the reasoning here. The TV's "gamma" function, whose exponent is nominally, say, 2.5, is supposed to (more than) fully offset the "gamma correction exponent" (say, 1/2.2) in the video camera. I don't quite see why that doesn't (more than) restore the original 1000:1 contrast ratio in the scene!

To be honest, I don't feel that Poynton is the best source in this regards. While he does state 1,000:1, later he seems to indicate that 100:1 simultaneous contrast ratio is sufficient in a display. This seems to contradict what he stated a few paragraphs before. In any case, it is quite clear that simultaneous CR above 100:1 is easily discernable even with relatively small viewing angles (such as smaller television monitors). In other words, I disagree with Poynton when he says that: "a simultaneous contrast ratio of 100:1 is adequate." In one sense, 100:1 is "adequate" for a decent image, but so is any other random and relatively low subjectively chosen simultaneous CR. His statement seems to imply that more than 100:1 simultaneous CR is not necessary, and that is not the case.

Quote:


Why? Mainly because nonlinear coding of TV signals (presumably including HDTV signals) reduces the specular highlights in video luminance by a factor of 10:1.

This is not directly from Poynton, and IMO is not the best reading. His discussion regarding the inability to reproduce highlight details from specular reflections refers to the luminances in a real scene, and not the much-reduced "tone altered" image captured on film or video. In essence, he is describing the limitations of a low-dynamic range capture and playback system and the inability to accurately reproduce the kinds of scene luminances (both in absolute and relative terms) in an image that we see in a real scene. The reduction in dynamic range through an LDR system (which basically includes all imaging systems in the history of humanity including drawing with a stick in dirt, except for the quite recent advent of HDR capture and display capabilities) is not from nonlinear coding. The same limitations would be present with linear coding as well, though of course the nature of the coding would be different.
post #218 of 505
Does Poynton give a reference for the 1000:1? I'd like to see that reference. Stating something does not mean it's true - as witnessed by hundreds of posts in this forum
post #219 of 505
In Digital Video and HDTV: Algorithms and Interfaces (actually, in the online errata PDF) Charles Poynton shows a diagram giving the luminance range of human vision. I have attached it to this post. It shows that human vision extends over eight orders of luminance magnitude, which is 100,000,000:1.

But the cone cells in the retina that enable color vision in daylight utilize only about the top five orders of magnitude: 100,000:1.

"At a particular state of adaptation," says Poynton, "vision can discern different luminances across about a 1000:1 range" (p. 197). So adaptation is like a 1000:1 elevator that goes up and down the 100,000,000:1 luminance range of human vision. In addition to bringing the cones into play in bright light, it brings the rods into play in dim light. (In intermediate lighting, apparently rods and cones both are active.)

Adaptation also opens and closes the eye's iris. Pupil diameter ranges between 2 mm and 8 mm.

Adaptation also "involves a photochemical process involving the visual pigment substance contained in the rods and the cones; it also involves neural mechanisms in the visual pathway" (p. 196).

***

Poynton (p. 196): "Total retinal illumination" is what mainly controls adaptation: "the mean luminance in your field of view. In a dark viewing environment, such as a cinema, the image itself controls adaptation."

My comment: if you watch TV in a dark room, the image itself again presumably controls adaptation. But with brighter ambient lighting conditions, the adaptation "elevator" goes up in response to the bright room lighting. Under such circumstances, it may filter out lower-level luminances in the TV image. Even if the TV's ANSI or simultaneous contrast ratio is, say, 1000:1, the low end of that range may be swallowed up by adaptation to ambient lighting.

***

Poynton (p. 196): "Dark adaptation, to low luminance, is slow: Adaptation from a bright sunlit day to the darkness of a cinema can take a few minutes. Adaptation to higher luminance is rapid but can be painful, as you may have experienced when walking out of the cinema back into the daylight."

My comment: A TV with a sequential or scene-by-scene contrast ratio (say, 10000:1) that is more than the 1000:1 the eye can handle at any given adaptation state will tend to hurt the eyes when it, the TV, goes abruptly from displaying a prolonged dark scene to a bright scene ... unless the contrast control is adjusted downward, sacrificing much of the 10000:1. If sequential contrast in excess of 1000:1 is thus squelched, simultaneous contrast will turn out to be far less, due to spill/stray light in the viewing room and flare in the display device. Poynton's norm of 100:1 for simultaneous contrast may be too low here, but figures like 300:1 sound reasonable.
LL
post #220 of 505
Eric: I think you should first read the entire thread, since these points have been made before (including the intrascene adaptation aspect, which I have mentioned repeatedly). Yes, the elevator moves across the entire luminance range, but most papers suggest that the "height" of the elevator is 100-300:1 rather than 1000:1. So Poynton's statement about 1000:1 without corroborating references is just that, a statement.
post #221 of 505
Quote:
Originally Posted by ChrisWiggles View Post

To be honest, I don't feel that Poynton is the best source in this regards. While he does state 1,000:1, later he seems to indicate that 100:1 simultaneous contrast ratio is sufficient in a display. This seems to contradict what he stated a few paragraphs before. In any case, it is quite clear that simultaneous CR above 100:1 is easily discernable even with relatively small viewing angles (such as smaller television monitors). In other words, I disagree with Poynton when he says that: "a simultaneous contrast ratio of 100:1 is adequate." In one sense, 100:1 is "adequate" for a decent image, but so is any other random and relatively low subjectively chosen simultaneous CR. His statement seems to imply that more than 100:1 simultaneous CR is not necessary, and that is not the case.



This is not directly from Poynton, and IMO is not the best reading. His discussion regarding the inability to reproduce highlight details from specular reflections refers to the luminances in a real scene, and not the much-reduced "tone altered" image captured on film or video. In essence, he is describing the limitations of a low-dynamic range capture and playback system and the inability to accurately reproduce the kinds of scene luminances (both in absolute and relative terms) in an image that we see in a real scene. The reduction in dynamic range through an LDR system (which basically includes all imaging systems in the history of humanity including drawing with a stick in dirt, except for the quite recent advent of HDR capture and display capabilities) is not from nonlinear coding. The same limitations would be present with linear coding as well, though of course the nature of the coding would be different.

Chris,

I assume LDR and HDR refer to low dynamic range and high dynamic range systems, right?

I think my reading of Poynton is correct, if to my mind a bit dubious in its logic. He seems to be saying (p. 83) that tone scale alteration of a video output signal is done by a video camera to:
  • compensate for the luminance of a TV screen being much lower than that in the original scene
  • compensate for the "surround effect" which lowers subjective contrast when viewing is done in dark or dim surrounds
  • limit the contrast ratio actually required of the TV, in view of spill/stray light from the TV and optical flare within the TV typically polluting the dark portions in the image
  • tamp down the luminance of specular highlights so they won't waste 90% of the luminance range of the TV

That fourth reason alone would reduce the usable simultaneous contrast ratio from, say, 1000:1 (the actual range the eye can accommodate in any particular state of adaptation) to 100:1.

If the sequential CR of the TV is set, via the contrast control, to above 1000:1, abrupt transitions to bright scenes can hurt the eye.

If the sequential CR of the TV is set to exactly 1000:1, then spill light, flare, etc. will lower simultaneous CR to a fraction of that.

If 90% of what remains is set aside for specular highlights up to 10 times brighter than diffuse white, for practical purposes the utilizable CR of the TV is squashed even more. That must be why specular highlights are crushed in the TV signal's encoding.

What I don't understand is why the TV's "gamma" doesn't un-crush them as it (more than) offsets the original "gamma correction" in the video camera. I have e-mailed Poynton to ask about that.
post #222 of 505
Quote:
Originally Posted by armadillo View Post

Eric: I think you should first read the entire thread, since these points have been made before (including the intrascene adaptation aspect, which I have mentioned repeatedly). Yes, the elevator moves across the entire luminance range, but most papers suggest that the "height" of the elevator is 100-300:1 rather than 1000:1. So Poynton's statement about 1000:1 without corroborating references is just that, a statement.

armadillo,

You are correct: Poynton gives no corroborating reference for his 1000:1 number (the supposed "height" of the eye's light-to-dark adaptation "elevator"). Also, it might be that the "elevator's height" varies with the level of adaptation. I'm going to read through the thread for references that contradict Poynton. Thanks.
post #223 of 505
Quote:
Originally Posted by epstewart View Post

I don't quite see why that doesn't (more than) restore the original 1000:1 contrast ratio in the scene!

http://www.cs.ubc.ca/~mmt/Papers/MscThesis.pdf
The most commonly used metric in LDR applications is the CIE 1976 standardization of lightness L*. L* is used in both the CIELAB and CIELUV color spaces, which target print and video respectively, and L* models contrasts approximately 100:1 and a peak luminance of somewhere around 200 cd/m2.

http://www.brucelindbloom.com/index....dCalcHelp.html
L* = 8 = .0089 (CIE 1976 lightness L*)
1 / .0089 ~ 112 CR

Optioelectric transfer functions (OETFs), such as the television standard Rec. 709 and the computer standard sRGB , which are similar to L*. A linear segment is included for practical reasons and the break occurs where the function equals an L* value of 8, corresponding to a contrast ratio of 100:1. Obtaining values below 8 is rare in practice and the break is considered the effective limit for video applications, reinforcing the fact that L* is only applicable to LDR images.

According to the law of Weber-Fechner we can roughly distinguish 1% difference of the background luminance level.

http://www.cis.rit.edu/people/facult...gACMTAPsub.pdf
Quote:


The human observer experiences a range of naturally occurring ambient light levels of nearly 14 log units and must be able to discriminate objects in the environment over 8 log units [Hood and Finkelstein 1986]. However, the differences in intensity reflected by those objects at any single light level are very small, spanning at best 2 to 3 log units [Walraven et al. 1990]

..or 100-1000:1
post #224 of 505
Quote:


but most papers suggest that the "height" of the elevator is 100-300:1 rather than 1000:1.

This has been repeated many times in this thread. Not a single paper has discussed this in the way you characterize it. I have not seen one paper state unambiguously that we can only see this small an amount across a scene.

A number of papers have been cited which quote figures significantly greater than this. This issue should not be tossed up to a preponderance of opinion, but thus far the preopnderance of that opinion as far as papers go clearly indicate that we are dealing with order of magnitude greater than 2.
post #225 of 505
Quote:


What I don't understand is why the TV's "gamma" doesn't un-crush them as it (more than) offsets the original "gamma correction" in the video camera.

Because the gamma is not really a direct part of the alteration of the tone scale. That is inherent in any LDR capture method, such as film. That was my point above, that he is not describing gamma as synonomous with "tone scale alteration" which is the way you seem to be interpreting it. This was what I suggested was a misinterpretation. Note that on p83 under the relevant section, gamma is not discussed at all. The following section on rendering intent discusses gamma, but mainly in terms of explaining why the end-to-end flow is not 1. For subjective reasons, and because we're using much lower luminances in displays and far smaller dynamic range than real life, just creating an image of a scene at this lower luminance and DR does not yield the best results, and some slight tweaking of gamma helps create a better image representation of the real scene. Note how crucially different this is than the complete implementation of gamma/de-gamma in video. He discusses the end-to-end exponent as 1.25 for television viewing. This is incorporated into the full implementation of gamma in video(by the fact that encode gamma is not the complete inverse of the assumed 2.5 display gamma), but is slightly separate. That is to say, film itself is not "gamma corrected" and then de-gammed upon display as video usually is. But here with film even without gamma/de-gamma being implemented at all, in a linear space, you are still left with an image that is going to be far smaller in DR than the scene. This kind of tone scale alteration is distinct from gamma/de-gamma in video, and is a natural part of any imaging system, including things like painting on canvas, drawing, etc.

So what you are confusing together are tone scale alteration and gamma. Tone scale alteration occurs inherently throughout the chain, and starts with the simple act of image acquisition from the scene. If you take any simple film camera, take a picture, and develop that film directly, you are not implementing any gamma-correction flow to the content per se. But the nature of the film and how it acquires the image naturally leads to a representation of the real scene that dramatically reduces the DR and alters the tones of the image, just by the very nature of film. Now, we can describe film as having a kind of response curve, it's often called an "s-shaped gamma" curve, which can be slightly confusing. But this is really quite different than the gamma/de-gamma flow in video. Gamma as it is implemented in video is mainly for signal/logistical reasons, and is not related directly to "tone scale alteration." In other words, if at the advent of video all displays were linear, we had digital signals, and we had plenty of data space and bandwidth, there would be little reason to implement gamma/de-gamma in the chain. But we definitely would still have a situation where images are drastically altered "tone maps" of the scene, just by the nature of the limitations of the reproduction system and the capture system. This kind of thing was happening hundreds of years before there was video at all (I'm no film historian, so maybe hundreds is an exaggeration!).
post #226 of 505
Quote:


"At a particular state of adaptation," says Poynton, "vision can discern different luminances across about a 1000:1 range" (p. 197). So adaptation is like a 1000:1 elevator that goes up and down the 100,000,000:1 luminance range of human vision. In addition to bringing the cones into play in bright light, it brings the rods into play in dim light. (In intermediate lighting, apparently rods and cones both are active.)

I agree with this characterization of this section of the text. What bothers me however, is that in the subsequent section he makes the reference to 100:1 simultaneous as being "adequate" for a display. This does not make sense, because even if we assume that he is previously correct with the 1,000:1 figure (in my readings and experience I think it's actually quite a bit higher than that across a normal spatially large scene), then it does not seem to make sense that a display with 1/10th that capability should be deemed "adequate" unless we are okay with mediocre capabilities. It does not seem to be the case either that he is confusing JND with black/white contrast ratios, as he's pretty clear at explaining JND/CSF stuff, bitdepth etc and keeping that distinct from his CR black/white discussions. That is why the 100:1 figure as "adequate" in the Contrast Ratio section strikes me as peculiar.
post #227 of 505
Quote:
Originally Posted by ChrisWiggles View Post

This has been repeated many times in this thread. Not a single paper has discussed this in the way you characterize it. I have not seen one paper state unambiguously that we can only see this small an amount across a scene.

A number of papers have been cited which quote figures significantly greater than this. This issue should not be tossed up to a preponderance of opinion, but thus far the preponderance of that opinion as far as papers go clearly indicate that we are dealing with order of magnitude greater than 2.

There seem to be so many scenarios for analyzing this. When you talk about seeing contrast across the screen, aren't we are dealing with seeing plus/minus CR around the APL? I had 1-200W lamp on in the living room last night. When I turned it off, I couldn't see to walk for a few seconds, then started to see more. While the eye is able to adapt to an extremely wide range of light to dark, you generally cannot see out of your house at night if the lights are on, and standing outside in daylight, you are limited to what you can see inside depending on the inside light level. You walk into a Theater, from the lighted lobby, and it takes time for you to be able to see in the dark. Not an extremely large CR.

Seems to me that when watching video, the eye will adjust to the APL and we can see contrast levels around that APL. If a full white field or very high APL image is on the screen for a while, when it suddenly drops toa a very low APL, or black, for an instant (seconds) we are blind until the eyes adapt.
post #228 of 505
Chris: I gave you several refs before. Here is one that handles everything from achromatic to selective chromatic contrast sensitivities both in the fovea and periphery. The highest sensitivities are for achromatic light of course and never exceed 130:1. Check it out yourself:

http://www.ncbi.nlm.nih.gov/entrez/q...=pubmed_DocSum

That is the abstract. Follow the link to PubMed Central to retrieve the original publication with all the graphs.
post #229 of 505
Quote:
Originally Posted by armadillo View Post

Chris: I gave you several refs before. Here is one that handles everything from achromatic to selective chromatic contrast sensitivities both in the fovea and periphery. The highest sensitivities are for achromatic light of course and never exceed 130:1. Check it out yourself:

http://www.ncbi.nlm.nih.gov/entrez/q...=pubmed_DocSum

That is the abstract. Follow the link to PubMed Central to retrieve the original publication with all the graphs.

I'm not sure I'm looking at the same article, because the article I am looking at doesn't cite this number anywhere that I see and doesn't have any bearing on the discussion at all.

I am looking at: "Human peripheral spatial resolution for achromatic and chromatic stimuli: limits imposed by optical and retinal factors" by: Anderson SJ, Mullen KT, Hess RF.

Is this the same article you're reading?

And if so, how on earth does one glean a contrast ratio figure of 130:1 as a maximum anywhere in that article? Please direct me to where in this article you think it is making this claim... I may have missed it, but I've looked through the article 3 times and I fail to see even the number 130 anywhere in the article.

First sentence of the abstract: 1. The aim of this study was to determine whether optical, receptoral or higher-order neural properties limit spatial resolution (acuity) in human vision, especially in the peripheral regions of the visual field. How is that relevant at all?
post #230 of 505
Quote:
Originally Posted by ChrisWiggles View Post

I'm not sure I'm looking at the same article, because the article I am looking at doesn't cite this number anywhere that I see and doesn't have any bearing on the discussion at all.

I am looking at: "Human peripheral spatial resolution for achromatic and chromatic stimuli: limits imposed by optical and retinal factors" by: Anderson SJ, Mullen KT, Hess RF.

Is this the same article you're reading?

And if so, how on earth does one glean a contrast ratio figure of 130:1 as a maximum anywhere in that article? Please direct me to where in this article you think it is making this claim... I may have missed it, but I've looked through the article 3 times and I fail to see even the number 130 anywhere in the article.

First sentence of the abstract: 1. The aim of this study was to determine whether optical, receptoral or higher-order neural properties limit spatial resolution (acuity) in human vision, especially in the peripheral regions of the visual field. How is that relevant at all?

armadillo thanks for the excellent and most germane reference. I have glanced through the paper and it will take some deciphering.

Of course ratios that you are used to seeing as a hobbyist would not appear in a paper such as this. I would guess that the 130:1 may derived from the foveal maximum at minimum spatial frequency used in the testing, it's going to take some noodling. Nonetheless, it would be helpful to understand how that ratio was derived based on this paper.
post #231 of 505
Chris: I can't really discuss the entire paper. Essentially, they used a moving sinusoidal grating (either achromatic, i.e. B&W or chromatic). They then assessed contrast sensitivity when the test person was able to detect the motion of the grating, meaning that they perceived conctrast. This was assessed both in the fovea as well as peripheral areas, where contrast detection is weaker. This produces the curves seen in the graphs. The contrast sensitivity is plotted on the y-axis (note that thes are log units). The contrast sensitivity is dependent on the frequency of the moving grating, the retinal area tested, and the chromatic information. The highest contrast sensitivities are found in the fovea for B&W patterns and this declines towards the periphery. Only green light has a similarly high contrast sensitivity as white light. Given that the y axis is log-based, the half-point between contrast sensitivities of 100 and 1000 corresponds to ~300. None of the test persons actually reach that half-point. So contrast ratios remain at 100-200. For colors it is 10-30.
post #232 of 505
Quote:
Originally Posted by armadillo View Post

Chris: I can't really discuss the entire paper. Essentially, they used a moving sinusoidal grating (either achromatic, i.e. B&W or chromatic). They then assessed contrast sensitivity when the test person was able to detect the motion of the grating, meaning that they perceived conctrast. This was assessed both in the fovea as well as peripheral areas, where contrast detection is weaker. This produces the curves seen in the graphs. The contrast sensitivity is plotted on the y-axis (note that thes are log units). The contrast sensitivity is dependent on the frequency of the moving grating, the retinal area tested, and the chromatic information. The highest contrast sensitivities are found in the fovea for B&W patterns and this declines towards the periphery. Only green light has a similarly high contrast sensitivity as white light. Given that the y axis is log-based, the half-point between contrast sensitivities of 100 and 1000 corresponds to ~300. None of the test persons actually reach that half-point. So contrast ratios remain at 100-200. For colors it is 10-30.

You must think we're stupid. Do you know what Contrast Sensitivity Function is? Try googling it. It has nothing at all to do with black/white contrast. It measures the ability to detect SMALL changes in contrast at a constant luminance, such as a gray. It also involves a spatial component, which is the frequency part.

Goodness, and here I thought we'd gone through all this before, and here we right back again with someone throwing around CSF figures as if they were at all germane.
post #233 of 505
Quote:
Originally Posted by mdtiberi View Post

armadillo thanks for the excellent and most germane reference. I have glanced through the paper and it will take some deciphering.

Of course ratios that you are used to seeing as a hobbyist would not appear in a paper such as this. I would guess that the 130:1 may derived from the foveal maximum at minimum spatial frequency used in the testing, it's going to take some noodling. Nonetheless, it would be helpful to understand how that ratio was derived based on this paper.

It's easy and exactly as I expected. A layperson saw a figure, saw a label that had the word "contrast" in it, and saw a number. You put those together, and voila!

And it's completely erroneous.

And incredibly hilarious that a number as precise as 130 was gleaned from those figures.


You can see why my patience for this kind of nonsensical sand-throwing has worn out. "look, look what I found! I found a number in some random PDF! WOW! I can't explain it to you at all, and I don't know what it means, except that you're wrong and I'm right! Hooray!"
post #234 of 505
Quote:
Originally Posted by armadillo View Post

Chris: I can't really discuss the entire paper. Essentially, they used a moving sinusoidal grating (either achromatic, i.e. B&W or chromatic). They then assessed contrast sensitivity when the test person was able to detect the motion of the grating, meaning that they perceived conctrast.

I haven't looked at the article, but if this is about the Contrast Sensitivity Function (CSF) then I have mentioned multiple times that this is a test of the least amount of CR a person can detect, not the most amount of CR. The CSF is a reciprocal of the contrast threshold. A score of 300:1 is a CR of something like 1.01:1. The approximately 300:1 goes down with age, which indicates that we need more CR as our eyes age to pick out the Just Noticable Differences, not less CR.

Just in case this wasn't clear, as far as the confusion with CSF that started long ago and still seems to be out there, imagine that the discussion was about whether you could tell if a black cat was super black or just pretty black as it was walking in front of a white wall and people kept bringing up testing of whether you could tell the difference between a white piece of paper and a white wall of just a barely different shade, as if the inverse of the percentage difference between the whites that are barely different would tell you the ratio for the white and black situation. Kind of like asking somebody what the biggest step they can take is and they start by testing what the smallest step they can take is and figure they can just take the inverse of that.

EDIT: I see that Chris covered it.

--Darin
post #235 of 505
Quote:
Originally Posted by ChrisWiggles View Post

It's easy and exactly as I expected. A layperson saw a figure, saw a label that had the word "contrast" in it, and saw a number. You put those together, and voila!

And it's completely erroneous.

And incredibly hilarious that a number as precise as 130 was gleaned from those figures.


You can see why my patience for this kind of nonsensical sand-throwing has worn out. "look, look what I found! I found a number in some random PDF! WOW! I can't explain it to you at all, and I don't know what it means, except that you're wrong and I'm right! Hooray!"

I do not beleive armadillo was aiming for a precise number but an estimation from a graph. Also not sure from my reading that the authors are referring to the CSF that Darin mentions

If you look at Figure 5 in the paper and further down in the text they mention that the limits of spatial resolution being ganglion density. They authors also define the periphery as > 25 degrees. From the graphs values and definitions of the x and y axes, 125-130:1 is a fair approximation for achromatic spatial acuity.

Oh and Chris, ease up a bit and read the paper. Perhaps some real anaylsis on your part could educate us all.
post #236 of 505
Quote:
Originally Posted by ChrisWiggles View Post

So what you are confusing together are tone scale alteration and gamma..... Now, we can describe film as having a kind of response curve, it's often called an "s-shaped gamma" curve, which can be slightly confusing. But this is really quite different than the gamma/de-gamma flow in video. Gamma as it is implemented in video is mainly for signal/logistical reasons

OMG

Chris it is you who has confused these two fundamental ingredients completely. Honestly you should research a given subject before lecturing anyone from a position of ignorance! Perhaps a google search could help?

http://www.sprawls.org/ppmi2/FILMCON/
"Conventional photography has also made use of gamma correction curves in the form of density-exposure curves, commonly known as D-log H curves in the art, by adjusting chemical balances in film and processing. This is analogous to the video use of gamma curves in that the film is now being used to alter the tone scale perception to create tone scales that are more pleasing to the human eye in a hard copy of the image"
Quote:
Originally Posted by ChrisWiggles View Post

you are not implementing any gamma-correction flow to the content per se..

EDIT/
The gamma correction, or contrast selection, is part of the photographic repertoire used to adjust the recorded image
post #237 of 505
Quote:
Originally Posted by ChrisWiggles View Post

I just did this in paint. The spread between squares isn't 20 pixels, that doesn't seem particularly important, it's roughly 40 pixels. The lowest square at 2% I cannot see even on an all black frame on an LCD monitor. I can on my CRT. So I made the dimmest square about 4% (level 10 in graphics range). I can see all four squares without any difficulty with the presence of the 100% square at the center. All squares are equal size, 250x250 on a black background. On a poor-ANSI display the presence of the white square may wash out and obscure the visibility of the other squares. But on an LCD which has pretty high ANSI, any person with reasonable visual acuity should have no difficulty at all seeing these squares at the same time.

I've attached the crude pattern I whipped up. It compressed down to jpeg to be attached, but that's what I'm looking at. (Note: attached image is for graphics range, NOT video range)


Also, I shouldn't have to note that this test is FAR less relevant than the hand shadow puppet tests described earlier because this test relies so heavily on ANSI CR performance of the display, making it basically irrelevant except as a measure of ANSI CR performance of a display. That we cannot see the dimmest square on many displays has little to do with our vision, and everything to do with poor CR performance of the display.

Hi Chris
I know I'm mad to get caught up in this thread
but...
I'm looking at your image that you created. I can see all 5 squares when I'm looking for them - my display clearly renders the differences, however when I look directly at the centre of the white square, the 4% square vanishes into the background. Also when I've stared at the white square for more than a few seconds, when I move back to have a look at the 4% square it takes a few seconds to see it again.
Do you see the same?
post #238 of 505
Chris: Sorry to frustrate you so much. You are right, the paper I cited is based on contrast modulation at equal luminance. I always thought that contrast modulation and luminance modulation were the same. According to the literature, this is apparently a matter of controversy. Not so much regarding the actual values, but regarding the mechanisms. Here is a link to a very recent paper that has assessed both and the differences are indeed relatively small:

http://journalofvision.org/6/4/3/article.aspx

The study concludes the following: "The results of the current study clearly show that there is no, or very little, difference in CE between LM and CM stimuli detection. This implies that observers are just as efficient for extracting LM signals from LM noise as they are for extracting CM signals from CM noise. Consequently, these results suggest that, after a second-order rectification, both stimulus types are processed by common mechanisms." CE = Computing Efficiency, LM/CM = Luminance/Contrast Modulation.

Finally, instead of getting all worked up about nothing, why don't you post a link to a scientific paper (not a web page or otherwise uncorroborated statements) that clearly demonstrates your numbers of 1000:1? Rather than discrediting my efforts to shed some light into the matter, why don't you as the world's foremost expert simply point us to the relevant paper and close the case?
post #239 of 505
Quote:
Originally Posted by ChrisWiggles View Post

Because the gamma is not really a direct part of the alteration of the tone scale. That is inherent in any LDR capture method, such as film. That was my point above, that he is not describing gamma as synonomous with "tone scale alteration" which is the way you seem to be interpreting it. This was what I suggested was a misinterpretation. Note that on p83 under the relevant section, gamma is not discussed at all. The following section on rendering intent discusses gamma, but mainly in terms of explaining why the end-to-end flow is not 1. For subjective reasons, and because we're using much lower luminances in displays and far smaller dynamic range than real life, just creating an image of a scene at this lower luminance and DR does not yield the best results, and some slight tweaking of gamma helps create a better image representation of the real scene. Note how crucially different this is than the complete implementation of gamma/de-gamma in video. He discusses the end-to-end exponent as 1.25 for television viewing. This is incorporated into the full implementation of gamma in video(by the fact that encode gamma is not the complete inverse of the assumed 2.5 display gamma), but is slightly separate. That is to say, film itself is not "gamma corrected" and then de-gammed upon display as video usually is. But here with film even without gamma/de-gamma being implemented at all, in a linear space, you are still left with an image that is going to be far smaller in DR than the scene. This kind of tone scale alteration is distinct from gamma/de-gamma in video, and is a natural part of any imaging system, including things like painting on canvas, drawing, etc.

So what you are confusing together are tone scale alteration and gamma. Tone scale alteration occurs inherently throughout the chain, and starts with the simple act of image acquisition from the scene. If you take any simple film camera, take a picture, and develop that film directly, you are not implementing any gamma-correction flow to the content per se. But the nature of the film and how it acquires the image naturally leads to a representation of the real scene that dramatically reduces the DR and alters the tones of the image, just by the very nature of film. Now, we can describe film as having a kind of response curve, it's often called an "s-shaped gamma" curve, which can be slightly confusing. But this is really quite different than the gamma/de-gamma flow in video. Gamma as it is implemented in video is mainly for signal/logistical reasons, and is not related directly to "tone scale alteration." In other words, if at the advent of video all displays were linear, we had digital signals, and we had plenty of data space and bandwidth, there would be little reason to implement gamma/de-gamma in the chain. But we definitely would still have a situation where images are drastically altered "tone maps" of the scene, just by the nature of the limitations of the reproduction system and the capture system. This kind of thing was happening hundreds of years before there was video at all (I'm no film historian, so maybe hundreds is an exaggeration!).


Chris,

I understand that there is a theoretical distinction between gamma correction and tone scale alteration, the latter being done to impose rendering intent. But Poynton says (and I believe) that as a practical matter the two are concatenated in a video camera's single, 0.5-exponent power function, used for encoding the television signal. In a standard display device a CRT a gamma exponent of 2.5 that is inherent in the electron gun offsets the original 0.5 encoding exponent ...

... but not completely. That way, the apparent contrast of the image on the TV screen is enhanced, as a way of compensating for oddities in how the human visual system perceives a TV image's low (compared to the original scene) luminance, when the image is being viewed in a dimly lit room (as opposed to the typical "bright surround" of a real-life scene).

What I still don't understand is how the need to compress specular highlights fits into this. Poynton says (p. 84) that the fact that they "must be compressed" in terms of their relative contrast by a ratio of roughly 10:1 is due to the fact that "at the reproduction device, we can seldom afford to reproduce diffuse white at merely 10% of the maximum luminance of the display, solely to exactly reproduce the luminance levels of the [specular] highlights!" He implies that the incorporation of rendering intent into the camera's gamma-correction power function the very next topic he is going to take up is what accomplishes (among other things) the 10:1 compression of specular highlights.

Yet it seems to me that the TV's gamma function with the assumed exponent of 2.5 automatically reverses most of that compression of specular highlights!

Perhaps the 0.5 encoding exponent compresses specular [i.e., non-diffuse] highlights in the scene by more than 10:1, and the TV's 2.5 decoding gamma then reduces the effective, overall highlight-compression ratio to approximately 10:1. Would that make sense?
post #240 of 505
armadillos paper that Chris had such as hard time with is indeed the CSF that Darin mentions as the least amount of detectable contrast. On page 3 under Methods they define what they mean by CSF as the reciprocal of threshold contrast. So if any so-called ratio can be derived from this paper with any confidence it would be the minimum not the maximum given the test parameters.

armadillos second paper is also less than helpful in determining maximum contrast detectability which is the topic of this thread. Although the other side of this argument hasn't supported their position outside of anecdotal evidence (Brightside paper doesn't count in my estimation since it's biased coming from a commercial company trying to sell product).

Bottom line here is if that anyone is going cite a paper tell us why something is or isn't so rather than just parroting the paper or posting a link.
New Posts  All Forums:Forum Nav:
  Return Home
AVS › AVS Forum › Display Devices › Digital Hi-End Projectors - $3,000+ USD MSRP › Some people are still being told that 200:1 is all we can see....