Perceptual Rendering Intent -- ProPhoto to sRGB

[cybis]

Photoshop action: Pro2sRGB_perceptual.atn
1-click action is available now.
Download the PS action from the below link.
 
Procedure
1. Load your ProPhoto image into CS.
2. Load PS action: Pro2sRGB_perceptual.atn
3. Shift F9 to execute the perceptual conversion from ProPhoto to sRGB

jc1, thanks for sharing! The last step of the action converts to the current RGB working space which might be set to something other than sRGB. Not sure if this was intentional. Shouldn't the last step be an assign sRGB?

I've only tried it on one image but was impressed by the extra details in deep blue shadows. Thanks!

.../Turns out BPC was turn off in my first convertion...

[jc1]

Hi,

The last step of the action converts to the current RGB working space which might be set to something other than sRGB. Not sure if this was intentional. Shouldn't the last step be an assign sRGB?

Yes, that was done intentionally.
As sRGB_D50_jc1 resembles sRGB (refer to Color Matching Test),  choice of assign function or convert function for this case, in my opinion, should has no impact on the final color reproduction. 

In my earlier post, I mentioned about higher ΔE and that might be due to colorimetric adaptation, when converting from aRGB to sRGB_D50_jc1. I shall investigate to see if there is any difference at all if assign is used instead of convert.

.../Turns out BPC was turn off in my first conversion...

Understood. The convert function has options such as BPC.
BPC is activated in Pro2sRGB_perceptual.atn.

Thanks for your feedback.

Regards,
jc

[cybis]

Yes, that was done intentionally.

Just to be sure, my RGB working space is set to AdobeRGB. When I run your action the final result is an image in AdobeRGB colorspace, not sRGB. Is that the intended result?

[cybis]

Understood. The convert function has options such as BPC.
BPC is activated in Pro2sRGB_perceptual.atn.

In the first two steps BPC is OFF. In the third step BPC in ON.

[jc1]

Just to be sure, my RGB working space is set to AdobeRGB. When I run your action the final result is an image in AdobeRGB colorspace, not sRGB. Is that the intended result?

Ooops! I see your point now.
If the working space is set to AdobeRGB, the final color space is converted (by default ?) to AdobeRGB  instead of sRGB. Let me check if CS is doing it correctly or assign function is more appropriate.

In the first two steps BPC is OFF. In the third step BPC in ON.

Yes, unsure about its final effect.

Regards,
jc

[jc1]

My bad. The 3rd step I set it to convert to AdobeRGB.
I shall post the update.


jc

[cybis]

Also currently, the first two steps are perceptual intent. I thought the first step should have been relative color, no?

[jc1]

Also currently, the first two steps are perceptual intent. I thought the first step should have been relative color, no?

Actually no difference in final result, as both are Absolute color space with no mapping function.


Pro2sRGB_perceptual.atn (revised)
To avoid further confusion, they are updated in the revised action to reflect actual intents used for the conversion.
For Step 1, rendering intent is set to Absolute, as both source space and destination space are with D50 illuminant.
The steps are shown below.



 
New download link for Pro2sRGB_perceptual.atn  <  revised >

As for jc1RGBR beta 2.1, no change was made, just a repost  < jc1RGB beta 2.1 >

Regards,
jc

[jc1]

Demo_ jc1RGB_Beta2.1 down load link:  < Demo_ jc1RGB_Beta2.1 >
 
It consists of following 6 files.
a. 0_jc1RGB (beta2.1).icc
b. 0_sRGB_D50_jc1 (beta2.1).icm
c. timage-50x100_prophoto.tif
d. jc1-saturated-color_prophoto.tif
e. Pro2sRGB_perceptual.atn
f. Pro2sRGB_perceptual_interactive.atn
 
 
Setup
1. Download and unzip Demo_ jc1RGB_Beta2.1.zip to a folder
2. Copy or Drag & Drop both 0_jc1RGB (beta2.1).icc and 0_sRGB_D50_jc1 (beta2.1).icm into system icc folder
3. Start CS and load both action files, Pro2sRGB_perceptual.atn and Pro2sRGB_perceptual_interactive.atn
 
note: PS action will remain in CS unless it is unloaded.
 
 
Interactive Mode
1. Start CS
2. Load one of above tiff files, or other image file with ProPhoto, BetaRGB or BestRGB profile.
    You may also like to conduct a torturing test with any file, be it sRGB or aRGB.
3. Press <shift>+<F10> to start interactive perceptual conversion.
    Follow the screen instruction or press Shift-C to continue.

            
 
4. Repeat by going to the drop down File menu, do a Revert.
 
 
Non-Interactive Mode
1. Start CS
2. Load tiff file
3. Press <shift>+<F9> for 1-step non-interactive conversion.
 
For further improvement, your feedback would be much appreciated.

Thank you.

Regards
jc
 
.../ Tips on PS action
.../ jc1RGB's Preliminary specification

[cybis]

I applied the 3 steps convertion to Bill Atkinson's twenty-eight balls test image and noticed some strange artifacts in the blue ball. Looks like banding (the Red value goes up and down several times).

[jc1]

I applied the 3 steps convertion to Bill Atkinson's twenty-eight balls test image and noticed some strange artifacts in the blue ball. Looks like banding (the Red value goes up and down several times).

Sound interesting.

I don't seem to be able to download this test image.
Could you advice where this test image can be downloaded, or pm me a link of that.

TIA

jc

[cybis]

Sound interesting.

I don't seem to be able to download this test image.
Could you advice where this test image can be downloaded, or pm me a link of that.

https://public.me.com/billatkinson Look for Twenty-Eight Balls.tif.zip.

[jc1]

https://public.me.com/billatkinson Look for Twenty-Eight Balls.tif.zip.

Thanks for the link.

Blue Ball Test

IMHO, examine color smoothness on screen is best by disabling the monitor profile, especially if monitor profile is created with cluts. As for monitor that is profiled with matrix, inaccuracy has direct influence on the color smoothness as well.

I agree with you that banding after conversion was observed, but unsure how severe it is, and whether or not it is within the norm. Other who has better experience may like to comment on that.

Here is my suggestion on how color smoothness should be examined on screen.
Set soft proof to monitor, zoom in 100% or higher, use the Navigator in CS to assist in viewing the problem area.

   


Following is not a direct apple-to-apple comparison, as one is converted from ProPhoto to photogamut profile and other from ProPhoto to sRGB, both did that perceptually.

   

I may carry out more extensive study on this matter.

Regards
jc

[jc1]

Blue Ball Test 2
 
The Blue Ball was extrated from the original file, Twenty-Eight Balls.tif which can be downloaded from https://public.me.com/billatkinson.

The RGB data pre and post conversion along with the original data points, were illustrated as follow.
a)  The converted data in Lab space do not deviate much between rendering with RelCol and with Perceptual.
b)  In xyz space (not shown), more shadow details were recovered with perceptual conversion, these additional details are likely from data of the original image outside the visible locus.
 

 
jc
 

[jc1]

The Mysteries of the 28 Balls

Photoshop CS5 vs Nikon Capture NX2

Following is one of them.

Observation
Differences in color space conversion between Photoshop CS5 (x32 dot4) and Capture NX2 (2.2.7).

Test point: Background color of the 28 balls, refer to below diagram for clarify.

Discrepancy
Refer to diagram, arrow 1
a) CS5 histogram, Pink line
b) NX2 histogram, Grey line  
 
    


                                        RGB             HSB            Lab
A) Direct RelCol

1 ProPhoto (CS5)            192,192,192     0,0,75         82,0,0
   ProPhoto (NX2)            192,192,192     0,0,75          ---

2 RelCol to sRGB    
    CS5 (ACE)                  203,203,203     0,0,80        82,0,0    
    CS5 (ICM)                  203,203,203     90,0,80       82,0,0
    NX2                          203,203,203      0,0,80           ---   

B) Perceptual workflow

1 ProPhoto (CS5)            192,192,192     0,0,75         82,0,0
   ProPhoto (NX2)            192,192,192     0,0,75         ---
-
2. AbsCol to jc1RGB
    CS5 (ACE)                  202,202,202     60,0,79       82,0,0            ACE throughout
    CS5 (ICM)                  202,202,202     30,0,79       82,0,0            ICM throughout
    NX2                           202,202,202      0,0,79          ---

3. Perceptual to sRGB_D50_jc1
    CS5 (ACE)                  203,204,203     144,0,80     82,0,0             Pink line?
    CS5 (ICM)                  204,205,206     193,1,81     82,-1,0
    NX2                           203,203,203      0,0,80          ---              Grey line?    

Notes:
All data were read directly from info screen.
Always on BPC
 
Comment
NX2 is consistence, but inconclusive.


Regards
jc

.../Further Observation
In CS, it seemed that the color space conversion was not affected by the problematic HSB values. With a color picker, I re-confirmed that the image was displayed correctly. Only the HSB info and possibly the histogram, are bad.

.../Clarification
Such phenomenon on histogram is probably due to very slight displacement on converted RGB values (R, G and B not exactly equal for grey) and it was further "amplified" on a compact info screen. So do not be alarmed by that. Similar effect was observed in NX2.

.../Logical reason I can think of now is smart PS modified the color of this line as it has the same color as the histogram background.

[jc1]

 
Effect of BPC
 
The effect on color space conversion due to BPC (Black Point Compensation) was studied with below 22-level grey scale.
 
 
 
Results (with Histograms)

                 

 
Comments
1. BPC has no effect on conversion with RelCol in absolute color space.
    In general, conversions in PS are carried out under condition D50 illuminant in that all absolute color spaces
    have the same black point ( Lab=0,0,0) and White point (Lab=100,0,0).
    Inviting comment.
2. For perceptual rendering, the effect due to BPC can be realized on shadow area.
3. There is no loss in  neutrality (not shown) and luminosity with perceptual conversion, in comparison to conversion with RelCol.
 
jc
 
 

[jc1]

Example: ProPhoto to sRGB Conversion without clipping on Highlight

  

Download link for the example's tiff image < Example1a >
 
Note: All my links are limited to 500 downloads.  Keep me informed if you are unable to download.
 
Thank you.
 
jc

.../update pic+link

[Peter_DL]

 
Here is a practical test.
Screenshots copy&pasted into a new file in PS, monitor profile assigned, Converted to sRGB.
Monitor comprises more saturated Reds than sRGB, so sRGB destination space is really the limiting factor.
Bpc enabled will all conversions:

Screenshot # 1:
Image a) at the left:  Straight conversion from Prohoto RGB -RelCol-> sRGB.
Image b)  ProPhoto RGB -Perc.-> PhotoGamut RGB -RelCol-> sRGB.
Image c)  same as the first one: straight conversion from Prohoto RGB -RelCol-> sRGB.
Image d)  at the right:  ProPhoto RGB -RelCol->  0_jc1RGB (beta 2.1) -Perc.-> 0_sRGB_D50_jc1 (beta 2.1) -RelCol-> sRGB

All quite close together so far, however, let’s do some post-processing on images b) to d) i.e. local-contrast-enhancement on the most saturated colors, as given with Screenshot # 2. I’ll desist from a final conclusion. Visually it seems that all other options are better than plain RelCol for this image.

Cheers!
Peter

--

[jc1]

Thank you, Peter.

Below are what I have observed.

Color de-saturation due to conversion with PhotoGamut RGB's is more obvious than with jc1's perceptual conversion. 
As for RGB clipping due to conversion, it would be nice if that can be shown with histogram.


Regards,
jc

[crames]

Example: ProPhoto to sRGB Conversion without clipping on Highlight

Download link for the example's tiff image < Example1 >

Although it has the ProphotoRGB profile embedded, the linked image does not seem to contain anything that exceeds the sRGB gamut. Is that the image you intended to show?