Using psychophysics (method of adjustment) to calibrate gamma based on JNDs - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 6 Old 07-16-2014, 11:09 PM - Thread Starter
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 793
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Using psychophysics (method of adjustment) to calibrate gamma based on JNDs

If and when the time comes where displays reach very deep blacks, I imagine it will be something of a challenge to profile their near black luminance responses. For those who don't want to rely on manufacturer's defaults, perhaps one solution would be to use a psychophysical approach to adjust the luminance.

Let's assume that the BT.1886 formula aims for perceptual uniformity, where the perceptual differences between each successive video level are uniform.

In that case, it would be a simple matter to write some code that allows the user to adjust the LUT value (of the neutral axis) to achieve this goal. It could work like this:

Display level 0 (min luminance). Display level 1, and adjust the LUT for level 1 until level 1 and 2 are barely distinguishable (the video levels could be separated in space or in time). Repeat all the way up to 255.

With a high contrast display, one might want two or three JNDs (just noticeable differences) between each video level. This would be more challenging, but it is feasible to do. You could do the above procedure for three successive levels (0,1,2), and then adjust 1 until it is indistinguishable from 2. And so on.

I may try to do this in Matlab using the psychtoolbox. If I do, I'll report the results - it will be very interesting to see how a perceptually uniform luminance function compares to BT.1886.

Any thoughts on this approach?
spacediver is online now  
Sponsored Links
Advertisement
 
post #2 of 6 Old 07-17-2014, 03:54 AM
AVS Addicted Member
 
Join Date: Sep 2002
Location: Wetumpka, AL
Posts: 15,343
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 103 Post(s)
Liked: 190
Seems like a lot of trouble once you figure in peculiarities of some displays, variability of content and variability of viewing environment.

With that said, it would be interesting to see if there's a worthwhile difference.

Samsung 64F8500, Panasonic 65VT50, Oppo 95, Tivo Roamio for OTA, Dish VIP722, Marantz AV8801 preamp, Rotel Amps, Atlantic Tech 8200 speakers, Seaton Submersive HP, Calman 5, Chromapure, Accupel DVG-5000, i1Display3pro, i1pro2, eecolor colorbox.
JimP is offline  
post #3 of 6 Old 07-17-2014, 06:06 AM
Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Posts: 121
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 27 Post(s)
Liked: 53
An interesting approach. A little difficult to apply to a computerized application. Forgive me for playing the devil's advocate but which demographics are you aiming at. Visual perception has been recorded to decrease as an individual ages. Even within a controlled demographic group the variance of visual perception is to great to pigeon hole.

The biggest issue with Gamma on displays of today is that it is trying to mimic displays of the past. This issue of legacy is not going to change. When manufacturers have their own ideas of how things should be done and then implement them, there is a lack of strict standards. This is where the calibration process is loosing the war. We strive to calibrate the Gamma within a -\+ 5% standard and the manufacturers produce displays that are -\+ 50 standard; it is self-evident that the issue of Gamma will not disappear in the near future.

I have noted by observing numerous calibration reports that others have submitted for one reason or another over the years. In the majority of the cases, no where are the target values of "Y" ever displayed and if they are, by no means do they appear to have been addressed. I have noted the "Y" values to be as much as 1000 points away from the target "Y". It is quite common to hear them ask why they have strange humps in their graphs. To say this once again, "Targets are their for a reason. If you don't use them how will know if you are on the mark?"

(Sorry about that, I will now get down off my soap box). Now back to the subject.

If psychophysics is is to determine the JND of a visual stimuli then one will have to standardize the point of subjective equality (PSE). As stated before the variables that exist when dealing with human subjects are vast. The problem would be to come up with a method of constant stimuli where by everyone could agree upon as a Gamma standard and then apply it to a system (as stated before) that appears to have loose standards. Again Gamma is somewhat subjective, Which Gamma are we using; 2.4, 2.2 or 2.0? So many variables ans so difficult to narrow down to a mathematical formula.

Please do not get me wrong, it sound like a very interesting idea and I am very interested in results that you may come up with; keep us informed.

Last edited by randal_r; 07-17-2014 at 06:20 AM.
randal_r is offline  
post #4 of 6 Old 07-17-2014, 06:31 AM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 591
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 50 Post(s)
Liked: 75
Quote:
Originally Posted by randal_r View Post
The biggest issue with Gamma on displays of today is that it is trying to mimic displays of the past.
You say this is a "problem", but no indication as to why. Merely being something from the past is not a reason.

The CRT EOTF was not an accident - it is plumb in the correct range to efficiently encode perceptually even light levels. Even starting with a clean sheet now, you would end up with a similar looking curve if tasked with the job of designing a means of efficient encoding, maximally resistant to noise or quantization.
gwgill is offline  
post #5 of 6 Old 07-17-2014, 10:04 AM - Thread Starter
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 793
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Quote:
Originally Posted by randal_r View Post
An interesting approach. A little difficult to apply to a computerized application. Forgive me for playing the devil's advocate but which demographics are you aiming at. Visual perception has been recorded to decrease as an individual ages. Even within a controlled demographic group the variance of visual perception is to great to pigeon hole.
The demographic would be the observer who is planning on using the display, in this case, myself

Yes, there is variance within the population, but that's a problem that manifests even if you stick to a certain standard, such as BT.1886.


Quote:
Originally Posted by randal_r View Post
The biggest issue with Gamma on displays of today is that it is trying to mimic displays of the past. This issue of legacy is not going to change. When manufacturers have their own ideas of how things should be done and then implement them, there is a lack of strict standards. This is where the calibration process is loosing the war. We strive to calibrate the Gamma within a -\+ 5% standard and the manufacturers produce displays that are -\+ 50 standard; it is self-evident that the issue of Gamma will not disappear in the near future.
I suspect that with the recent efforts to characterize human response to changes in luminance all the way up into the higher ranges of luminance (I believe there is work being done by Dolby and others in this area), there will be a push on the industry side of things to produce tighter native luminance functions on displays. This with respect to the upcoming UHD standards, of which high dynamic range is a component (and it's pretty clear that BT.1886 will not be appropriate on high dynamic range displays, especially with only 8 bits).

Quote:
Originally Posted by randal_r View Post
If psychophysics is is to determine the JND of a visual stimuli then one will have to standardize the point of subjective equality (PSE). As stated before the variables that exist when dealing with human subjects are vast. The problem would be to come up with a method of constant stimuli where by everyone could agree upon as a Gamma standard and then apply it to a system (as stated before) that appears to have loose standards. Again Gamma is somewhat subjective, Which Gamma are we using; 2.4, 2.2 or 2.0? So many variables ans so difficult to narrow down to a mathematical formula.
The resulting "gamma" will be determined by the results of the experiment. And I was proposing method of adjustment, not method of constant stimuli (the latter is certainly feasible, and might produce more accurate results, but would take quite a long time).

Quote:
Originally Posted by randal_r View Post
Please do not get me wrong, it sound like a very interesting idea and I am very interested in results that you may come up with; keep us informed.
I certainly will!
spacediver is online now  
post #6 of 6 Old 07-17-2014, 10:29 PM - Thread Starter
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 793
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Ok, so I just wanted to get a sense of how responsive my system is to changes in the LUT, so I did an experiment with Matlab and Psychtoolbox (similar to what I imagine dispcal -R does)

Starting with a linearized gamma table, I displayed a full screen image of a light grey patch.

The gamma table was 256 rows, each specified with a precision of 0.0001, scaled between 0 and 1.

I then incremented the values in the table by increments of 0.0001, while measuring this grey patch with the i1 display 3 (running off a nearby laptop).

I was able to get consistent changes in luminance with LUT value increments of 0.0005. For example, if I raised the LUT by 0.0001, the luminance reading didn't budge (or rather, it didn't budge from its fluctuation of about plus/minus 0.01 nits). But as soon as I raised the LUT by another four bumps of 0.0001, the reading suddenly jumped up.

For example, at gray level 100 (out of 255), the reading was 4.50-4.52 nits.

The 100th LUT value before any adjustments = 0.3882.

I changed the value to 0.3883, no change in luminance whatsoever, however, when the value changed to 0.3887, the luminance suddenly went up to 4.54-4.56 nits.

If I'm interpreting this correctly, that means that changing the LUT by five ten thousandths had an impact on the voltage being sent through the DVI-VGA cable. This is a precision of one part in two thousand, which is roughly 11 bits.

Next, I just have to figure out a good psychophysical approach to estimating JNDs (I'll look through the literature), but it's good to know I have at least 10 bits of LUT precision to work with.

Last edited by spacediver; 07-17-2014 at 10:44 PM.
spacediver is online now  
Reply Display Calibration

User Tag List

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


Forum Jump: 

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

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