Page 1 of 1

Profile conversion to sRGB - avoid blocked shadows?

Posted: August 27th, 2013, 8:28 pm
by Charles2
What is your technique for converting to sRGB?

Frequently when I convert from AdobeRGB (my working color space) to sRGB, shadows block. Attached is an example showing the HSV-V histograms. (The screen capture is in sRGB, so the image on the left in AdobeRGB is not displayed well.) I converted with Maintain Full Gamut, but the other choices also caused the shadows to block.
PWPprofile_conv_MFG.jpg
PWPprofile_conv_MFG.jpg (61.93 KiB) Viewed 8799 times

Re: Profile conversion to sRGB - avoid blocked shadows?

Posted: August 28th, 2013, 3:00 pm
by den
Subscribing to a 'does it look right' artistic sense and what one wishes to convey, here are several ways to unblock an image's shadows regardless of cause while preventing change to the low midtones to highlights tone range...

Method 1:
...Extract the HSV-H and HSV-S channels of the sRGB image version and Combine them with the Extract-ed HSV-V channel of the aRGB image version...

Method 2:
...create a 1.8 gamma image version of the 2.2 gamma sRGB image, then using a 0 to 15% shadow tone range mask, Composite-Blend to a preference Amount, the 1.8 gamma image shadows into the 2.2 gamma sRGB image while monitoring Preview...

Method 3:
...use a RGB BrightnessCurve Smooth curve [0,0], [50,65], [100,100] with a 0 to 15% shadow tone range mask with the sRGB image as input, adjust the white slider Amount to a preference while monitoring Preview...

Method 4:
...use the sRGB image version as Input with a 0 to 15% shadow tone range mask and subtractively filter it with itself to a preference white slider Amount while monitoring Preview using the Filter transform adjusted to a preference ExposureCompensation setting...

Addl thoughts...

Suggested Method 1 occurs in the HSV color space model where the sRGB hues and saturations are unchanged with increasing shadow tone range Value... ...so existing shadow range chroma noise may become more visible.

Suggested Methods 2,3, and 4 occur in the RGB color space model, so as shadow Luminance is increased, shadow chroma noise will tend to loose saturation, perhaps becoming less noticeable.

An assymetrial 0-15% shadow tone range mask for 3456x2304 images I tend to use, can be formed by:
...1) click on the sRGB image and open the Mask Tool
...2) lower-left Apply Add a Brightness BrokenLine curve: [0,100], [5,100], [15,0], [100,0]
...3) lower-left Apply Add a Blur = 15
...4) lower-left Apply Subtract a Brightness BrokenLine curve: [0,0], [15,0], [35,100], [100,100]
...5) lower-left Apply a +2 Feather
...6) lower-left Apply a -2 Feather
...7) leave the Mask Tool 'active' on the sRGB image

...den...

Thanks, den, a cornucopia as usual

Posted: August 28th, 2013, 4:38 pm
by Charles2
Thanks, den, a cornucopia of techniques from you as usual.

Method 1a:
Perform method 1; Composite | Lighten the unaltered sRGB with the result through a mask of lowest 15% shadows.

Re: Profile conversion to sRGB - avoid blocked shadows?

Posted: August 29th, 2013, 5:17 pm
by den
Charles2...

(1) I am not so sure that if one processes image data starting with raw conversion in one color profile and finalizing it to preference nuances [tone, colors, contrasts, and sharpness] in that color profile that a simple conversion to a differing color profile will result in the retention of the nuances without additional processing to restore preferences... ...perhaps the best way to prevent shadow blocking when converting to sRGB is to initially raw convert to that profile and process to preferences within the sRGB properties.

(2) Something I have considered but have not investigated... ...what happens to the aRGB image version if it is first linearized [changed to a gamma 1.0] before the full gamut profile conversion to sRGB with perhaps that resulting image's gamma restored to 2.2?

There are instances where (1) gamma corrected re-sampling [re-sizing] and (2) gamma corrected sharpening have preference nuance retention that are not otherwise retained when transformed at a non-normalized gammas... ...perhaps a color profile conversion would behave similarily.

...den...

Re: Profile conversion to sRGB - avoid blocked shadows?

Posted: August 29th, 2013, 5:29 pm
by Charles2
den wrote: I am not so sure that if one processes image data starting with raw conversion in one color profile and finalizing it to preference nuances [tone, colors, contrasts, and sharpness] in that color profile that a simple conversion to a differing color profile will result in the retention of the nuances without additional processing to restore preferences... ...perhaps the best way to prevent shadow blocking when converting to sRGB is to initially raw convert to that profile and process to preferences within the sRGB properties.
Well, how much trouble is sRGB worth? It is just that sometimes when converting to sRGB, noticeable harshness appears. It may be blocked shadows in one image, too much brightening in the highlights in another. Lost details are lost, so thanks again for the solutions that borrow from the aRGB image.

Re: Profile conversion to sRGB - avoid blocked shadows?

Posted: August 30th, 2013, 1:16 pm
by tomczak
This link (and the link within it) may shed some light on why it happens: I believe it's the linear section of sRGB gamma curve that is responsible for it.

I have tried to fix the problem by making a special sRGB (by editing it) with a single normal gamma curve (i.e. no linear section at the deep shadows). The way it was supposed to work was to convert the AdobeRGB to this special sRGB (profile and data) and than just changing the profile name (without changing data) to regular sRGB. I recall it worked, but I haven't used it much since.

http://www.dl-c.com/board/viewtopic.php ... srgb#p3298