P: Histogram behaviors are different from prior versions

[FeedbackCommunityMember]

I am experiencing problems with the histogram in 19.1.0. I am a mid-career fine art photographer by profession, and have been a PS user since Version 2. I am also a member of the Authors Guild, and write on photography for various publications. I have been purposely hanging back at PS 2015.5.1, as it has served me well. Yesterday, I decided enough, is enough, and installed 19.1.0.

This first histogram is at Cache Level 1 for the file values of a 36Kx24K pixel 16-bit grayscale file.

 

Please notice how there are "tails" (lines) out each side of the main body of data, indicating that there are small levels of data almost to the limits of range. For me, it is important to know about these tails exist so that I do not end up creating a clipped condition when applying a curve function. We use S-curve limiters to compact the tails without clipping.

Here is the exact same file at Cache Level 1 for the same 36Kx24K pixel 16-bit grayscale file, but this time in 19.1.0.

 

Please notice how there is no tail indicating data extending to the left, and rather a botched one going to the right. This is not helpful! You will also notice that the Mean and Standard Deviation values differ.

Further, it use to be nice to be able to take the cursor and scan across the histogram with a display of level and count showing up for whatever was under he cursor. This seems to have gone away in 2015.5, and is even worse in 19.1.0.

Thanks for your help.

Pete

[FeedbackCommunityMember]

I don't believe testing with real data is appropriate as we don't know what should be showing in any part of the histogram.  Here is a file to test:  https://www.dropbox.com/s/3h02hzeyw2xsara/grayFlat512gray16.tif?dl=0 and the display of the histogram from my system.    This file has four squares of data.  each is colored so it is separate from the others.  The large square does contain the other three.  The histogram is scaled by the area in the large square minus that of the three smaller ones.



All of the live bins should be different height but the same intensity.  Viewing in levels or curves has the same problem.  This is not the same type of display as an image where edge effects etc are accepted.  This is a diagnostic and it should display the ground truth.  Photoshop is used to do cosmetic changes to photos but it still should be maintained on a scientific level.   

RONC

[FeedbackCommunityMember]

I agree Ron. We need to be able to use tools in PS that give us scientific precision, not just approximations for cosmetic betterment of our photos. LightRoom can take on that other voice. Further, PS should not be a "design" platform either---it should be about precision imaging. Just like LightRoom, I think the "design" elements need to be taken out of PS proper, and have their own platform. PS needs to be about imaging, imaging, and imaging.

And if floating point is too complicated to maintain views in histograms, then 32-bit fixed point would be logical. I think jumping forward to floating point backfired horribly.

I can understand LightRoom on an iPad. I can understand a design version of PS on an iPad, but I sure can't understand why PS proper needs to be on an iPad. It does not lend itself to precision imaging.

The Curves function and the Histogram function in PS are archaic. The concepts are correct, but the tools are so small and limited that one cannot get the precision they need in a 16-bit world. These simply look like the same concepts that were produced before PS Version 7. Competitors are showing vastly improved histogram and curves functions, but they are not complete because of the lack of precision and save curve functions---a saving opportunity for Adobe to get it done right before they are leapfrogged.

Also, just to understand, I am doing math processes that are as wide a 200,000 pixels on my photos. More testing needs to be done before release to make sure that PS is operating correctly with these large file sizes. Every time I run into a 30K pixel limitation, I wonder why we are in the dark ages.

Similarly, there shouldn't be anything in PS that can't be run at 16-bit, and sadly this is not the case. That should have been resolved post Version 7.

I was working on making complex split tones of my monochrome works using color gradients, and it finally dawned on me that the color graduations were throwing 8-bit data against my 16-bit monochrome images---not 16-bit X 16-bit. The concept was great, but thwarted by internal limitations in PS.

[FeedbackCommunityMember]

I believe that diagnostics should be at a scientific standard in all Adobe programs and operating systems.  I also believe that the backbone of all Adobe programs should be 32 bit floating point or higher.  Most of the problems occur because the coder is not familiar with what the tool is used for.  If you work on histograms your should find out how Adobe's clients use histograms.   I'm afraid Adobe management doesn't see what needs to be maintained and not. 
I don't think there are 30k pixel limits in PS if from Adobe but most third  party older plugins are always suspect.  Floating point was a very poor kludge thinking it was for HD type work only and not the whole image business.  
I would say the top management at Adobe is so consumed with how much they make that it is a moot point.
 Enough philosophy.
RONC

[Todd Shaner]

Gentlemen, this post is quickly going down the rabbit hole! One of the requisites of this forum is to keep requests to one specific issue or new feature. Since the disappearing curves issue has been solved that leaves us with the Histogram accuracy. Please use a separate post to discuss these other issues.

Using *Ronald Chambers file the histogram looks identical to his screenshot on my system. *Pete Myers what are you seeing? *Jeffrey Tranberry what are you seeing?

"All of the live bins should be different height but the same intensity."

If you calculate the height as a percentage of the 512x512 area that is mid-gray (117) the 64x64 area is 1.6%. On my system the Histogram is 123 px in height. .016 x 123 px = 1.968 or about 2 px, which is what I see displayed. The 32x32 area is .40% or .004 x 123 px = .49 px (less than 1 px). On my display it shows as one pixel because that as low as it can go and still be visible correct? The 16x16 area is .10% or .001 x 123 px = .123 px. So how do you show these differences using 1 px? Adobe has done this by by lowering the brightness. It makes sense to me and appears to be working properly on my system.


Having said that *Pete Myers histogram using *Jeffrey Tranberry file shows NO tails, but both my and Jeff's system show the tails. Please scroll up to see for yourself. No need to copy & paste them here! Clearly something is different with Pete's systems when using a non-scientific image file. That is in fact what *Pete Myer complained about...poofing histograms aside! *Ronald Chambers do you see the tails using Jeff's image file?

[FeedbackCommunityMember]

I see some "tails" to either side but I don't know whether I'm seeing what I should. I see different from what Pete is seeing. Again you have to use some image where you know what is supposed to be there.
RONC

[Todd Shaner]

"I see some "tails" to either side but I don't know whether I'm seeing what I should. I see different from what Pete is seeing."

That's the issue. On Pete's system the tails are not visible at all, which obviously isn't correct! Pete can you please download Ronald's test file and show us the histogram?

 https://www.dropbox.com/s/3h02hzeyw2xsara/grayFlat512gray16.tif?dl=0

As per my analysis using your scientific test image file it appears to be correctly showing the relative levels within the constraint of the small histogram palette. If I were to request a change it would be to offer an option to allow the user to resize the histogram palette to make it larger.

[FeedbackCommunityMember]

First off, I apologize for going off tangent.  I need an attitude adjustment and hope it is coming soon..

Attached is a display of histograms from Jeff's image the left is from full size and the right from a about 1/3 size version.  Notice the difference in the tails and I can't guarantee there aren't ones still missing.  I think allowing the histogram to be physically larger should help but there needs to be special treatment so that a bin of count 1 is fully visible. 

I'm going to try a couple things with a plugin I have.  Usually the bottom level is only half the interval so rounding up needs to be done.  Artificially add one to non-zero bin could help.  Problem is the maximum bin is setting the criteria.  Non-linear (log) display is often used to show a different view but it will have the same problem with very large images.

 

Width of input to histogram is 3014 vs 1024 and height scaled likewise.  

RONC

[FeedbackCommunityMember]

Correction: there were three issues interrelated, and my other post got merged to this one, so I would assume that we will be addressing all three in this thread. One is resolved:

1) The Histogram tails not showing, which for me has been an issue with every version of PS past 2015.5. Note: two different computer systems and OS to go with that, so I doubt it is as simple as an OS problem. However, as Ron has demonstrated, there seem to be hidden setup items in PS that I may have missed post 2015.5. The histogram tails issue is not a new problem for me, but an ongoing conundrum.

Also, you are showing me tails, but those tails need to match the data. I cannot tell from the histograms posted alone if you are showing data from Level 1 at a Count of 10, to Level 246 at a Count of 01 on Jeff's photo. You asked to match data on a known photo, but the histogram is not the whole story---so please include the data.

And is that Windows or Mac OS? Which version and which PS version are you using Todd?

2) The intercept line is missing from the curves function.

3) RESOLVED: the histogram function disappearing on touching the curves function (or any other function).

So let us not forget the intercept line as it too is an important part of the Curves function.

Pete

PS Sorry for the breakout into other issues related, but darn if there isnt a lot left hanging! How many posts do we have to make to have some forward focus on the broad issue of improving the histogram and curves function for use in the 21st Century? And in my thinking, having these other issues come up is directly linked to Adobe not keeping their eyes on the fries. I think that is exactly what Ron was trying to say as well---am I correct Ron? It IS related.

[FeedbackCommunityMember]

I'd ask how many Adobe executives can really use Photoshop or other programs.  Most companies are weighted toward sales types and bean counters at the top.   Enough.

RONC

[FeedbackCommunityMember]

Hi Todd:

I don't participate in the Adobe forums very often. Are you the moderator of this forum or a participant? I am not sure what "champion" is, so forgive my asking.

Pete

[FeedbackCommunityMember]

Hi Jeff:

Thanks for your interest there at Adobe. I hope I have provided enough information to understand the two remaining issues---the intercept line and the histogram tails. If I can be of further help, please search for me via Google, and be in touch direct.

Else, signing clear.

Pete Myers
Santa Fe

[FeedbackCommunityMember]

Jeff/Todd,
I would like to share my recommendations on how to make histograms more usable.  If yes, let me know. 
RONC 

[Todd Shaner]

Ron, the tail extension in the resized 1024 width image file is most likely due to Bicubic interpolation ringing artifacts. Lacking any other evidence the histogram is most likely correctly displaying these artifacts.

https://en.wikipedia.org/wiki/Bicubic_interpolation#Use_in_computer_graphics

[Todd Shaner]

1) What's causing the "missing" histogram tails on your systems remains to be determined and what I was trying to focus on. Unfortunately this post ths become very fragmented with numerous posts concerning other issues, suggestions and observations on how to "improve" the histogram.

2) The intercept line issue has been entered in a separate post and acknowledged by Adobe with status of 'In Progress.' 

3) Disappearing histogram SOLVED. Now let's focus on the "missing tails" issue.

https://feedback.photoshop.com/photoshop_family/topics/intersection-line-in-curves-not-working

[Todd Shaner]

The Champion title just means Adobe has recognized the individual's contributions and product knowledge and offered ability to communicate directly with Adobe Product Management. We also have the ability to moderate posts and report issues.

I am also an Adobe Community Professional (ACP), which is a separate program that provides the ability to work more closely with Adobe Forum and Engineering staff including prerelease testing.

https://community.adobe.com/t5/user/viewprofilepage/user-id/9983072

[Todd Shaner]

The best way to do that is by creating a separate post here in the Photoshop Family forum. Please keep it specific and focused on improving the PS histogram(s). Once completed please copy the URL link to that post back here in a reply. That way people landing here can add their comments and 'Me To' vote on your post, which helps to get Adobe's attention. Adobe Company Admin *Jeffrey Tranberry has replied here and is following this post.

If you come up with any ideas on how to fix the histogram "missing tails" on *Pete Myers systems please let us know in a reply back here. It's clearly a separate issue unrelated to the histogram accuracy question.

[Todd Shaner]

Pete, in order to investigate the remaining open issue concerning missing histogram tails can you please provide information for the two system models you are using including OS version, GPU model, and display model, including actual resolution settings if different than the display's native resolution, and OS scaling setting if other than 100%. Since the missing histogram tails are only 1 pixel in size I suspect display path interpolation may be the cause for their disappearance.

[J453]

2. Yep. I was able to reproduce the issue with the intersection line not showing, so that's in engineering's hands.

We're down to trying having a file/setup to reproduce the tails issue #1.

[J453]

Yep. Todd is a kind helper/moderator and he meets with Rikk and I on a regular basis to discuss issues and the products.

[J453]

You'd be surprised. A lot of executives are hardcore users of our products.

[FeedbackCommunityMember]

I'll put together a separate post to talk about enhancements to the histogram displays.  Is there any other option other than Histogram, Curves, Levels. and ACR where the Histogram is displayed?  This should be carried across all Adobe programs with histogram displays.

I can't duplicate Pete's missing tails so expect it is a parameter not a bug but like the disappearing histogram before I remembered the collapse to icon option, I can't think of what.

RONC

[Todd Shaner]

Histogram, Curves, Levels are the only places I can think of that would benefit from improvements for better accuracy.

[FeedbackCommunityMember]

In playing with Jeff's image, I find that there are no "Zero" count bins so every bin position should show something.  I will show this and more when I post in thread.  I think that there a problems with the present Histogram but this does not explain what Pete is showing us.

RONC

[FeedbackCommunityMember]

I was incorrect about there being no zero count bins.  

But something more disconcerting is that I mad the Jeff's file RGB (each is same as the gray value).  I then ran the Histogram on the 16 bit RGB image.  See attached:

On the left is a Histogram from my plugin and it shows live data from the low level to about 8 bins from the high level edge.  This is what the gray histogram should look like.  On the right are the output of the latest PS version run on the RGB image.  There are the parts missing as before in the discussion but look at the blue one.  It is much like Pete's example.

There seems to be a bug in the RGB version which also might show up for the grayscale one. 
 

Please try doing this yourselves.  I didn't measure the height of each histogram but the bottom one might a pix smaller.

My version is what I'm going to purpose when I've viewed enough data to feel secure in it.
RONC

[Todd Shaner]

Good catch! I can confirm what you are seeing in the All Channels view of Jeff's file after selecting Image> Mode> RGB with ProPhoto RGB working space. The Blue channel has no tail and stops at about level 222. the Red and Green channel tails stop at about level 246. So definitely somethings wrong and the root cause is probably related to what's causing all of the tails to show missing on Pete's system.