Color space conversion, color details and banding.

[bbd10]

When we develop and edit picture starting from RAW then very often we can lose fine color details during conversion to final color space. The typical remedy is to define working color space wide enough to accommodate wide camera gamut, but still, our final color space should be smaller so finally, a conversion is unavoidable. Leaving aside a discussion about what is real color, for  sometimes complex reasons we can allow small color shifts and keep details visible. Often happens that color banding or unpleasant color areas emerge in final result even when you use 16 bits per channel and perfect profiles.

 

To overcome this problem you can try to use wide gamut profile for working space, use the color proof layer on top and then export  or convert picture to final color space.

 

Below is the explanation (rationale) for this procedure.

 

For example, let's imagine that there is pixel with R:G:B values of 116:51:32 with ProPhoto D65 profile. For simplicity, I will use here 8-bit encoding. Normally, with wide gamut profiles, at least 16 bits per channel should be used. Now, when we convert an image with this pixel to sRGB its value becomes 164:10:27, but color will not change. Why? Because this particular color is inside sRGB and ProPhoto gamut. When I will use color proof correction layer in ProPhoto profiled picture, then for this particular pixel value will not change, and still will be 116:51:32. Similarly, there will almost be no change in expected value after conversion to sRGB color space. Because of a rounding errors small shift to 163:11:26 can be observed. This particular pixel value is almost on the boundary of sRGB color space, so small changes are possible.

 

The situation will change dramatically if we will try to do the same with a pixel which color is out of sRGB gamut. Let's take saturated red pixel in ProPhoto color space, where the value will be 163:0:0. Now, with proof layer applied, pixel value should change to 170:72:26, but when you directly export this picture to sRGB space then the value will be 241:0:0. Exporting picture with proof layer applied results in 241:8:5. Why there is such difference? The reason is simple: proof copy is still in ProPhoto space, so relative colorimetric routine adjusts not only saturation but also other parameters. When you directly export to small, sRGB space then conversion routine is in a completely different situation. Oversimplifying, zero is zero, so closest pixel value is  for conversion routine 241:0:0 - a distance between a pixel value in different colorspaces is simply too large.

 

This behavior leads to "fine tuned" color space conversion routine. This routine gives good results with highly saturated pictures and with RAW workflow, where there are many color details outside sRGB (aRGB)  space. You have to remember, that modern cameras have really wide gamuts. Very often much wider than Adobe RGB in reds and blues.

 

What do you think?