Re: Choir: Dan's comments


Dan Margulis
 



On Mar 19, 2021, at 5:08 PM, Doug Schafer <k_d@...> wrote:

Dan, I appreciate your answer and it drives me to ask more...although I "think" I know the answer.

To achieve  “individual acts of unusual cleverness” to remain, I would think if the multiple images averaging was not "stupid" (i.e. pure equal balanced %'s) but rather "smart/intelligent" where user decides how much % for each layer to place emphasis on keeping certain effects, Then sometimes an intelligent par version could be better than stupid par?

Presumably it’s *always* better, but often the gain is so small that it wouldn’t be worth the extra time in thinking/execution.


And if true, are there any special "tricks" or techniques? Or simply a matter of trial/error (intelligently done) as to layer stacking order and % applied from each layer.

It’s usually one extra layer at a time. We are rarely in a situation where we have five versions to play with. Use a second version to improve a first, and then a third if necessary.

For example you often suggest a blend of 25% or 40% etc to get the desired effects?

The exact percentage would be a matter of taste. The more important question is what blending mode to use and whether to employ a mask.

And I assume we must use a background/bottom layer of 100%?

Yes.

And, for the choir image, if you did a different intelligent par image, which images to blend and what order, and what percentages? 

I don’t want to go back and plan it out, but I’m looking at the five that currently reside in my tentative par folder for the Pagodas. I’m not going to go any further than looking because I assume that some of these will be replaced by better versions. But if I had to work with these, calling them #1-5 in the order I’m looking at them:

First, there are two that correspond more or less to my own personal preference and therefore are, in my opinion, the two best. They’re so alike that it’s unlikely one can improve the other. So I might blend them 50-50 or I might just compare and pick the one I prefer and toss the other. Either way I’m down to four versions. Let’s say I’ve decided to discard #5.

Then, I find the one that is most unlike my base version, which presumably should be the final blend. It happens to be #2. I note also that its color is very good but I’m not so satisfied with its luminosity.

#3 is not a favorite but it does have one interesting feature. I’m not sure there’s an advantage to using it.

#4 is unusual, it has a very nice sky, the foreground is about as good as the others, but all cool colors are the wrong hue.

So my presumed order of attack is:

*Discard #5
*Put #4 on top of #1 in Luminosity mode to find out whether I really need the color of the sky. If not, change mode to Luminosity and use one of the RGB composites as a layer mask so as to emphasize the excellent sky. If yes, I’ll use Normal mode for the layer but after applying the RGB to the mask I’ll apply the red in Darken mode, to further restrict the cool colors in #4.
*Put #3 on top and see if there’s any point in doing anything with it.
*Put #2 on top in Color mode, because I’m sure I want more of its color than its detail. Adjust opacity to taste.
*Put another copy of #2 on top in Luminosity mode and adjust opacity to something less.

I expect that this would produce something a bit more attractive than blending all five at 20% each, but I wouldn’t expect a difference as big as there would have been in the Choir image.

I could see wasting many hours and achieving nothing....so where to start? How to think thru an approach? Pick the best where one has a problem offset by another? Or simply pick the best 5 and trial/error blend them?

I definitely wouldn’t look for the best five, but rather the five that were most useful for blending. I’d look for (or create) ones that counteract any presumed deficiency in what we have. For example, if we like our first attempt but suspect that it might be too cold then we create a second version heavy on the orange, knowing that if it doesn’t turn out to be useful we can trash it.

Dan



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