Re: Niagara Spray--comments on individual versions.


Dan Margulis
 



On Jun 10, 2020, at 9:52 AM, ROBIN MARK D'ROZARIO <rdrozario@...> wrote:

Two questions out of this:
1) Is False Profile/Multiply superior to Screen/Multiply or is it user/photo specific.

To some extent it’s image-specific but basically FP (or Exposure)/Multiply is more flexible and less dangerous.

If you look at these entries you’ll find several where the lower half of the face looks nice but the top half, around the eyes/eyebrows/forehead is too dark. The chances are that these are the people who used Screen mode.

Screening an overly dark face definitely adds contrast to it because lighter areas lighten much more rapidly than relatively darker ones do. With FP the face actually loses contrast temporarily, it becomes closer to the lighter areas (in our case, closer to the weight of the background falls) and overall the image is temporarily too light.

Tonal contrast in a face is sometimes a good thing and at other times not. I would prefer not to pre-empt the decision of whether to add it by using Screen while still in RGB. I’d wish to postpone the reckoning for another day, preferably when I’m in LAB.

Screening also may not lighten the image enough, and you may or may not have to make some kind of artificial mask. With FP/Multiply at least eight ready-made masks are available; by default I’d use the RGB composite from the background layer unless the dark areas are considered more important than the light ones, in which case I’d use RGB/Merged layer. There are also six individual channels that might make a better mask as well. Here, for example, I’d be looking for a mask that was light in the waterfall and darker in the face. That would be the green channel, where the contrast between these two areas would be greater than in the RGB composite.

If handled carefully all methods work, though some people make them more complicated than others. The PPW panel carries 1.0 and 1.4 gamma false profiles. The false profile, artificially lightening the image, is needed because otherwise the multiplication would probably plug the shadows, mask or not. The only time I use anything other than 1.0 and 1.4 is when the image already is somewhat light and there is no danger of shadow plugging. In that case I sometimes assign ColorMatch RGB (which is 1.8 gamma, rather than the 2.2 most of us use) and multiply away.

It was interesting to note, however, that many of the people who assigned false profiles in the Niagara exercise said they were using some gamma value other than the above.

2) Is there a permanent way to deal with the purple/Magenta blotches that magically appeared in this iphone image?  I assume they are an issue in other highly processed smartphone images as well.

They are indeed an issue at least in iPhone captures. In my experience they affect color only, not darkness, so they can be fixed directly in the A channel of LAB. Although the B channel is home to these defects as well it is not nearly as pronounced as the A so it can usually be ignored. 

These blotches show up  in areas of little color variation. Apparently Apple’s engineers have decided to save on storage and computation by ignoring minor AB variations. The area around the magenta splotches is a splotch as well, basically no variation in the channel at all, we just don’t recognize it as a splotch because it’s the color we expect.

If they needed to save on resource use, I tend to agree with them: these splotches can offend a professional audience but I doubt that the viewers of the Niagara Spray image would even notice. So I rarely bother taking them out.

There are several ways to fix them, some more complicated than others. If a similar blotch for some reason appeared in a capture from a better camera, there wouldn’t be much alternative to working directly on the A channel. But with the A and B in these areas being so detail-free in iPhone captures, the following kloodge generally works:

1. In LAB, make a duplicate layer

2. Make a rough feathered selection around the blotch, keeping the blotch to less than half the selection area.

3. Filter: Blur>Average.

4. Change layer mode to Color and adjust opacity if needed.

Again, this is unlikely to work with captures from a higher-quality source.

Dan

Join colortheory@groups.io to automatically receive all group messages.