Images devoid of actual depths - unusual behavior with PerfectRAW

[robyferrero]

Something that's never happened to me before: When I opened this converted file with MakeTiff, the PerfectRAW response curve was quite abnormal. Then, after working on it, I managed to fix it as shown in the attached example.
However, I wanted to understand why a file that appears perfectly normal to me had its initial curve so drastically altered.
As I said, it doesn't happen with other files, so I assume it's not a problem with my Color Perfect. In any case, if anyone wants to download the RAF file to see if the same thing happens, you can download both the RAF and the Linear TIFF file here.

[C.Oldendorf]

Roberto, howdy,

thank you for this illustrative image post. I enjoyed looking at its hidden details.
I think I can explain what is happening here, and in this particular image the explanation is actually rather simple.

The darkest meaningful parts of the scene are probably the black feathers of the two gulls. But those feathers are not truly black, and at the same time they make up only a tiny fraction of the entire image. So the algorithm behind BP Tails, which lets us define a certain percentage or fraction of a percent of the image that is to become true RGB 0/0/0 black, has very little suitable information to work with. In such a case it can run astray, and with it the associated definition of the black point colors, meaning the relative difference between the three channels when the black point is set.

This is not a bug in ColorPerfect, but a case where this particular setting is unsuitable for this particular image. With digital images, the solution is often simply to scroll BP Tails down until the image behaves naturally again. In more extreme cases, it can be better to turn BP Tails off completely.

I am attaching a close-up photograph of a pink rose from around twenty years ago that shows the same principle very clearly. That image also has no real depths. The truest color is obtained with the black point system turned off entirely. If BP Tails is scrolled down to its minimum value, the result can still be photographically useful, and if that is the rendering the photographer wants, there is nothing wrong with using it. Strictly speaking, however, it is no longer fully true to the scene. If you slowly scroll BP Tails upward through the whole range in the rose image, you can see the BP Tails algorithm going crazy as well.

For images like these, it can actually be better to work with White and Graded White rather than relying on the black point system itself. There simply is no substantial true-black information in the picture for BP Tails to lock onto reliably.

Your cliffside image in bright sun, with only the two small gulls providing the darkest values, is much the same kind of case. It is a beautiful image, by the way — just not one that naturally contains the kind of deep shadow information that this specific black point setting expects except at very low settings. The turning point seems to occur around 0.000009!

There is also a slightly funny aside. When checking this, I tried interpolation through MakeTiff using both LibRaw and Adobe DNG Converter. With Fuji X-Trans files, LibRaw introduces interpolation artifacts that are dark enough to mask this particular problem. Of course, that does not make the image better in quality terms, but it means one might never have noticed this peculiarity in the first place.

[robyferrero]

Yes, after this explanation, everything is clearer. Now that I know, if I have to think about how a base curve works, your explanation is enlightening.
In fact, it was easy to get to the result I achieved with the two seagulls, just as it was easy to balance the rose in question precisely because I learned the importance of BP and BPTails.