Odd transparency between layers

Shooting in macro mode, techniques, tips & tricks
Post Reply
Posts: 1
Joined: 15.12.2019 04:50

Odd transparency between layers

Post by themann42 »

I'm trying to get a clean stack and running into a little issue where intersecting edges overlap. I started with Method B, and the overlapping areas have a blurry edge running the length of the overlap. I then tried Method C, and the overlapping areas don't have a blurred edge, but the program seems to have a difficult time deciding which layer should be on top, and actually just blends them together creating an odd transparency between the two areas.

Just wondering if anybody more experienced to this knows of a way to get a cleaner stack that would have clean edges with no transparency. Perhaps something in Retouching (I haven't been able to really understand the Retouching section yet). I tried a few things with no success. Left to my own devices with no solution, I would use Method B and then photoshop the blurred edges away (not ideal because there's a lot of overlapping areas). Thoughts?
Screen Shot 2019-12-16 at 2.39.27 PM.png
Method B
(3.72 MiB) Not downloaded yet
Screen Shot 2019-12-16 at 2.37.57 PM.png
Method C
(3.53 MiB) Not downloaded yet
User avatar
Posts: 597
Joined: 29.04.2019 22:38

Re: Odd transparency between layers

Post by Catherine »

Most probably this can't be solved through adjustment of rendering parameters. We'd recommend using method С and retouching the intersection area.
User avatar
Posts: 34
Joined: 12.12.2019 08:26

Re: Odd transparency between layers

Post by umbel »

themann42 wrote: 16.12.2019 23:47Thoughts?
What Catherine said. The other commonly used technique is subslabbing/substacking, which is more of a Zerene thing as Helicon stack splitting doesn't produce aligned substacks. Stas has suggested including a consistent first image for generating aligned substacks in Helicon but this is tedious to do manually and requires the substacks have sufficient depth connectivity. Substacks tend to work best with bristly insect type geometries and aren't as effective with the smaller depth separations of plants.

I commonly deal with hairs similar to the ones in your images and retouching pyramids (method C) onto depth maps (method B) followed by F9 source image cleanup is the fastest method I know. No stacker I've tried (which is most of them) does exceptionally well with this sort of depth reconstruction but, by adjusting lighting and diffusion, I think you might be able to obtain somewhat improved SOOS (straight out of stacker) results.

Algorithmically, an obvious next step is to automate pyramid to depth map retouching and the process of reviewing the stack against the source frames for correct object ordering in depth and object continuity. The retouching process is basically another contrast maximization problem similar to weighted averages, depth mapping, and pyramids. The source frame checking is an object recognition and continuity thing. So the pieces of a machine vision assist exist. But I'm not aware of any code which attempts to assemble them.
Post Reply