# Photoshop: Lab mode increase precision

• Idea
• Updated 2 years ago
• (Edited)
Photoshop CC:I would like to have greater precision for Lab values. Today they are only integers even with 16bit / channel. Also 32bit/ channel does not support Lab. This lack of precision even effects RGB888 values where L:0to100,a:-80to80,b:-80to80. Perhaps a future update can address this?
• 4 Posts

Posted 5 years ago

• 20280 Posts
CIE LAB is designed such that a very good observer, under very good lighting conditions, can just distinguish a difference of 1.0 between adjacent colors in any dimension (L, A, or B). Decimal places for LAB does not make much sense unless you are doing further processing on physical measurements or some scientific imaging.

32 bit/channel is for high dynamic range imaging, and LAB has no useful definition outside of unit range on L, and about +-128 on AB. So LAB in 32 bit/channel doesn't make too much sense unless you were doing scientific imaging and know how to limit LAB to the appropriate ranges.

Photoshop 8 bit channel LAB offers steps of 0.4 in L (0..255 mapped to 0..100), and 1 in AB (directly mapped to -128..127).
Photoshop 16 bit/channel LAB offers steps of 0.003 in L (0..32768 mapped to 0..100) and 0.0078 in AB (-16384..16383 mapped to -128..127).

Why exactly do you think you need more precision in LAB? (the last time someone wanted this, they were actually trying to work around a horrible print profile, and rebuilding the profile worked a LOT better)
• 199 Posts
I'm finding a need for greater precision particulrarly when working with paper white simulations. When taking measurements with a spectro and translating those Lab values to the RGB working space the rounding is significant and this is especially seen in the highlights. Here's what you can to do see this for yourself: In the Color Picker, start with L=100 a=0 b=0. If you decrease the "a" value by one integer the R value drops 2 points. Increase the "b" value by one integer and the B value drops by two points. While this may not seem visually significant to you, for color science projects like G7 Certification this increases our DeltaE variances to unacceptable levels. Why not allow us to define Lab values to at least one decimal point? For color science professionals this would really make a difference.
• 199 Posts
Bump - any comments on this?
• 4 Posts
I am using photoshop for scientific imaging.

Given,
Quote:
Photoshop 8 bit channel LAB offers steps of 0.4 in L (0..255 mapped to 0..100), and 1 in AB (directly mapped to -128..127).
Photoshop 16 bit/channel LAB offers steps of 0.003 in L (0..32768 mapped to 0..100) and 0.0078 in AB (-16384..16383 mapped to -128..127).

On a color picker / entry dialog (whether it be 8/16/32 bits per channel) it makes sense that I would be allowed to enter the appropriate number of significant digits to whatever color space I choose to maximize the accuracy of the my entry to the available bits per channel does it not? I am afforded this in RGB why not the other color spaces?

I imagine this would be useful for other as well.
• 20280 Posts
Well, the color picker is more for visual and qualitative color selection, except in the case of HDR where more precision and range really is necessary.

Yes, there are requests to offer greater precision in the pickers, but when tested with users it never works the way they imagine and gets voted down.

I have done some LAB work by just relabeling the channels as RGB and using 32 bit/channel --- but I know how to deal with the limits of LAB, and don't need a visual preview of the results until after I convert back to 16 bit/channel.

Maybe it would help if you explained more about what you were trying to do?
• 4 Posts
I create color targets for camera and display testing. I also create printed and display targets to test the limits of human visual perception.

I could certainly do the conversions by hand for the color picker and many of the things I do I do with scripting so its not hard to do the math there either. I could use and sometimes do use other tools such as Matlab but I just happen to like the interface of Photoshop. It just seems to me that Photoshop could be a bit more exact with the colorpicker and the different color spaces, I'd also like to see the same precision in the info window.

Thanks.
• 20280 Posts
Yes, they could expose 16 bit/channel values. Though some people might get even more confused with the 0..32768 or -16384..16383 ranges.

You can already show 16 bit and 32 bit/channel values in the info palette -- just select 16 or 32 bit values for the readout.
• 4 Posts