Focus Merge: Potential Improvement

[yusha]

Hi,

Please compare the first two attached items, which are the final outputs of a focus stack.

1+2+3+4:  VS  :A+B (i.e. [1+2] + [3+4])

Here's what's going on:

1+2+3+4.png is a focus stacked image from 1.jpg, 2.jpg, 3.jpg, & 4.jpg >> The result is blurry, so I tried dividing the task to help the alignment algorithm.

A (1+2).png is a focus stack using only photo 1 & 2, while B (3+4).png is one using 3 & 4. >> These 2 intermediates, when combined will contain all 4 focus points (as would 1+2+3+4).

A+B.png is a focus stack of A (1+2).png and B (3+4).png. >> This result is what I expected initially when merging the 4 photos at one go.

This means that the current alignment is not optimal, perhaps you could offer an alternate mode where Photo would proceed by using intermediate image states, as I did above.
There are different ways to approach this, perhaps the same result could be obtained by first combining 1 and 2, then the resulting image to 3, and finally the latter combination to 4 to get the final output. Or alternately, use some kind of algorithm/AI to determine the closest set of images to merge sequentially before the final render.

Is this logic legit? I tried, it's repeatable.

Thanks.

1:     2:     3:     4:     A (1+2):     B (3+4):

Source_Photos.zip

Edited August 17, 2017 by yusha
Attached an archive so one can download the source files easily to test out.