Restacking (ie, stacking stacks)

[dgarnick]

Once I get my settings right, I often shoot several identical stacks. Then I compare the output from my stacks, and keep the best. However, I've noticed that sometimes one stack will be the sharpest in one area, and another stack will be the sharpest in another area. Out of curiosity I tried stacking the outputs. The results were great! I got the sharpest part of each output merged into one image.

BTW, this also made me a bit skeptical of objective/tube tests that rely on only one stack for each objective/tube. The difference between stacks was definitely noticeable, albeit, at the pixel peeping level.

[rjlittlefield]

It seems completely reasonable that stacking again would collect the best of all bits, given sharpness differences in otherwise identical intermediate results,

But I am curious about the origin of differences in the first place. Have you tried to identify what is causing them?

--Rik

[dgarnick]

Well, it's observable when I pixel peep. I'm assuming it's vibration; only some shots in a stack show slight blur - so each stack is likely to have different slices blurred. I'm pretty conservative on the focus increments. I have a heavy table and a setup of about 15 lbs of aluminum (t-slot and bracing) on anti-vibration feet. I take the photos late in the evening when there's no traffic on the road in front of my house. The floors do transmit vibration all too well. I shoot mirror-up and use a 1 second delay between shots - maybe I should up that to 2 seconds. And I might get a stone slab.

My main motivation for posting was how successful restacking is.

I'm probably obsessing about the sharpness. However, when I do artistic work, I do heavy post processing that emphasizes any blur. And, my prints are typically about 2 feet across - sometimes larger.

[rjlittlefield]

Thanks for the further information. That all makes sense. BTW, pixel-peeping obsessions are not unusual around here. I think this tendency has been responsible for a lot of advancements over the years.

The re-stacking approach is interesting and I am not sure what the tradeoffs are. Speaking as the fellow who wrote Zerene Stacker, I am concerned about the possibility that slight errors in alignment will accumulate differently in two nominally identical stacks, leading to misalignment artifacts when the re-stacking is done. But that will be very different from one sort of subject to another, so it's a try-it-and-see sort of thing.

An alternative approach that addresses the same issue (occasional blur due to vibration) is to shoot with a smaller focus step, so that you get the same number of source frames in only one pass through the subject. Then you have the options of either (A) treating the source as several stacks by skipping frames, using Options > Preferences > Preprocessing > "Stack every N'th frame" and loading images starting with frames 1, 2, etc., or (B) treating the source as a single stack that includes all the frames at the same time.

There are good arguments for why approach (B) might be expected to give a better result, particularly at high magnification where the "squirming around" effect can be significant. It will also result in slightly less defocus blur in areas between focus planes, even in the best case where vibration and squirming are not in play.

--Rik

[Lou Jost]

There are good arguments for why approach (B) might be expected to give a better result, particularly at high magnification where the "squirming around" effect can be significant. It will also result in slightly less defocus blur in areas between focus planes, even in the best case where vibration and squirming are not in play.

I think the advantages of B would be lost if the vibrations were typical stochastic environmental vibrations like those of cars and trucks, which last more than a few seconds and will ruin multiple shots. In that case, stacking stacks should be better, as the ruined sets will usually be in different parts of each stack.

[dgarnick]

Thanks Rik and Lou for the suggestions. I'll experiment.