One Way to fix a Flower Shot

[xpatUSA]

Hello Flower-Shooters,

I have a shot of a purple flower that does not convert well from camera space into sRGB:



In another thread, a discussion had started about color saturation, lightness and the CIELAB color space.

Saturation and Exposure - busting a Myth

Based on that discussion, I tried editing the flower shot in RawTherapee's 'Lab' space and was pleasantly surprised to find that the smudgy signs of over-saturation in the right-hand image were removed just by increasing the Lightness L*.

Indeed, I watched the histogram in awe as the previously-suppressed greens at the left of the histogram marched firmly back into the picture! I smiled too, as some contrast started to appear in those areas which were pretty smudgy before.

I played a little more with Lightness and also the Chromaticity sliders but left the Contrast slider as-is. Then I sharpened it up a bit. IMNSHO, the result is that the image at left is clearly superior to that at right.

Lab space processing and deconvolution sharpening: don't leave home without them

[DanK]

Interesting post. thanks. I'll play with this more. I have been using the lab color space for other reasons, but not for this.

[mikesan]

I too am a great fan of the tools available in RawTherapee. The Lab tools seem to have been very effective in this image.

I am curious to know how you conclude that the loss of detail/contrast in the 'before' image resulted from 'over-saturation' as opposed to simply under-exposure. The problem is clearly corrected by using the L*. What happens if you use only the Chromaticity controls?

[xpatUSA]

I am curious to know how you conclude that the loss of detail/contrast in the 'before' image resulted from 'over-saturation' as opposed to simply under-exposure.

Yes, the conclusion came from a long experience of converting Sigma X3F files in their converter, Sigma Photo Pro (SPP). In other words, the conclusion did not derive from just this one image. If, during conversion from camera raw, any one channel gets forced down to zero, the resulting RGB color is, by definition, 100% saturated in the world of HSB. The extent of this forcing depends on the target color space which for me is the smallest - always sRGB - and therefore more likely. In SPP, the method for an immediate fix is to a) ensure that the working space is your target space and then b) reduce saturation until the "zero'd" channel shows a positive value. The result is almost identical to what I saw in RawTherapee when increasing L*.

The problem is clearly corrected by using the L*. What happens if you use only the Chromaticity controls?

The chromaticity control affects the magnitude of the a*,b* vector. If that vector starts outside of a color space's gamut, reducing it's value will bring it inside of that gamut -at which point, it is no longer over-saturated. Not unlike reducing saturation in SPP.

Hope this helps . .

[xpatUSA]

The chromaticity control affects the magnitude of the a*,b* vector. If that vector starts outside of a color space's gamut, reducing it's value will bring it inside of that gamut -at which point, it is no longer over-saturated. Not unlike reducing saturation in SPP.

Hope this helps . .

Maybe that didn't help some folks, so I've illustrated the above with an example:



Here we see the CIELAB color space with the a* and b* axes shown. Let's say you're looking at a color in RawTherapee's native ProPhoto color space and let's say that color is the blob marked "OOG". Also we see the gamuts of two popular color spaces aRGB and sRGB and that the blob is outside of them both, hence it is indeed Out-Of-Gamut.

Now we see none of the above in RawTherapee's review page and anyway the blob only represents one pixel. Still, the effect of the Chromaticity slider is more easily understood in this diagram. When the slider is adjusted negatively, the blob will slide along the yellow line in the direction of the arrow. A certain amount of slider adjustment would bring the blob to within the aRGB gamut. Further adjustment would bring it within the sRGB gamut. The point at which the blob passes each gamut boundary is when the color becomes less than 100% saturated for that color space. Going to the limit, if the blob reaches the center, it has no color at all because there, a* and b* are zero. All it has is Lightness, which is not part of this discussion.

I know I shouldn't write "aRGB" but it suits the diagram and the description . . .

[IzzieK]

I am taking notes, Ted....I will read this again off line and sink it in slowly in my brain. Thanks for the explanation.

[mikesan]

Ted,
Thanks for the detailed explanation. This stuff is mostly above my pay grade, but in the correction you demonstrate in post #1, you correct the oversaturation using the L* slider. How does that affect the OOG colors?

[xpatUSA]

Ted,
Thanks for the detailed explanation. This stuff is mostly above my pay grade, but in the correction you demonstrate in post #1, you correct the oversaturation using the L* slider. How does that affect the OOG colors?

Oh boy, that's a tough question. Have you ever looked at a color space gamut in full, by which I mean viewed in 3D as a volume?

A typical space has it's wider part toward the top of the volume. Lower down (darker), the space is much narrower. Thus, by "moving" a color upward, i.e. toward lighter, it can go into the gamut volume from underneath - at which point it is no longer OOG, by definition.

A poor analogy might be a helicopter, rising vertically, and entering a cloud?

[xpatUSA]

Oh boy, that's a tough question. Have you ever looked at a color space gamut in full, by which I mean viewed in 3D as a volume?

A typical space has it's wider part toward the top of the volume. Lower down (darker), the space is much narrower. Thus, by "moving" a color upward, i.e. toward lighter, it can go into the gamut volume from underneath - at which point it is no longer OOG, by definition.

A poor analogy might be a helicopter, rising vertically, and entering a cloud?

I guess my last was a bit of a threadstopper, Mike

So, I've done another diagram which might help mo' better:



Here we see the gamut volume of the sRGB color space, viewed more or less from the magenta "corner" of the gamut. Underneath, we also see the widest parts of the gamut projected downward and shown at the L*=0 level. Any color at that level is, by definition, out-of-gamut because L*=0 is pure black and can not have a color, just as RGB = 0,0,0 has no color. I've shown purple at the same level and it is obviously out of the sRGB gamut being way below the gamut volume. Now, if we increase the value of L* for that purple while leaving the values of a* and b* the same, it can be see that it will rise upward toward the gamut volume and meet it where the arrowhead is. Any further upward and it will enter the gamut volume and become literally in-gamut.

The 3D CIELAB representation is the very best I know of for a good understanding of this whole subject

Anything else is pure obfuscation, IMNSHO