OK, this is looking better.
Thanks for sending me the full-resolution version offline. Let me walk you through a few issues that I see here and that you asked about separately.
First, the alignment issues that you saw earlier were almost certainly due to reduced size images (1000 pixels wide). The earlier version of the code essentially assumed camera-resolution images and was sometimes inaccurate with smaller ones. That problem should be fixed in the current version.
The ghostly halos are typical of the so-called "pyramid" methods found in ZS, CZP, and TuFuse. They are essentially negative after-images of flower parts from out-of-focus frames. There is a corresponding brightening of areas around the dark parts, but that is usually harder to see.
With PMax (and corresponding methods in all the other tools), there is also an issue with contrast buildup. When a feature is seen completely focused, and slightly OOF, and moderately OOF, and extremely OOF, that feature in the final image ends up receiving brightness contributions from multiple levels of focus. Those contributions add up so that in the final image, the feature has more contrast than it did in any of the original images. When detail is low contrast to begin with, the buildup is often a good thing. But when the detail is high contrast, the buildup can cause clipping at black or white. It can also cause a sort of "glowing" effect.
DMap does not have these issues, but it has different ones. DMap is prone to a different kind of halo, and rather than enhancing contrast it tends to overlook low contrast detail altogether. With a deep stack and low contrast detail, DMap (and corresponding methods in all the other tools) is vulnerable to what I call "stacking mush", in which certain areas lose all detail even though the human can easily find frames with detail in them.
Despite its simple appearance, this test subject you have chosen is actually quite challenging. It has intense contrast against the background, which makes halos very likely, yet it has low contrast over much of the petals.
In my experience, this sort of subject cannot be handled really well by either DMap or PMax, or any other single method or fully automated combination that I know of in other tools.
However, it should be fairly straightforward to handle by manual retouching, using a combination of DMap and PMax as outlined in the ZS documentation. This is exactly the strategy that I used with the
mayfly face that I recently posted. I've used it on many other stacks also, but not talked much about it.
In your specific case, retouching is problematic because the retouching tool takes even more memory than PMax, and it seems your 1 GB is not quite enough. To get some practice with it, you will need to reduce your image sizes. That will also help with speed issues that might crop up in using large images and correspondingly large brushes on a slow machine.
You asked in email about the apparent transparency of some flower parts, especially on the left. That problem commonly occurs when shooting a deep stack at wide aperture. It is caused by the lens "looking around" OOF foreground structures and seeing detail in the background. The software has no good idea what is foreground or background, and no idea at all of what is opaque versus transparent, so it just renders out everything it sees. The only fix for this problem is manual retouching. With a deep stack, it can be helpful to render the whole stack, and one or a few substacks that cover just the foreground structures, then retouch from the substack output images into the full stack output.
Hope this helps. Glad to hear we're making progress on making this thing work.
--Rik
) I'm aware now that I'd inadvertently left the camera on "Vivid" which has not helped.
4x, OK I'll try that.
) Rik spends a lot of time patiently repeating stuff, he's done it for me enough times!
.
.
in 7 min, using PMax in the newest ZS version (Thanks Rik!). The resulting 58 MB, 16 bit image is posted in the Technical and Studio Photography -- Macro and Close-up forum.