Looking for feedback for a new (Windows) focus stacking software

Hi,

I am running a beta of a new focus stacking software for Windows (named FocusFusion), and I am looking for feedback. It is available in the Microsoft Store, and details can be found on my website as well (https://www.delphitools.info/focusfusion/)

It supports 48bit TIFF, regular JPG and JXR, with two algorithms, which are processed at 16bit per channel:

• a classic Depthmap, meant for general use, with a required ordering or shots (back to front or front to back)
• a wavelet algorithm for low contrast focus stacking (wavelet algorithm can be very noisy otherwise).

There is a basic retouching tool included at the moment (a clone brush). The processing is meant to be as fast as possible without requiring a GPU.
DNG support is currently limited to Windows WIC, so while this can work with DNGs out of DxO PureRaw, this won't work with "raw" DNGs.
I've only tested on my own shots, with my own workflow, so I'm looking for less "own"  feedback

dhhddre wrote:

Hi,

I am running a beta of a new focus stacking software for Windows (named FocusFusion),

What would this bring to the table to differentiate it from Helicon and/or Zerene?

I tried it on 2 scenes, each having 7 images. Wow, I was impressed. I will continue using my Zerene program but your program is terrific for the 2 landscapes I used it on. I did not try using its editing tools. My only suggestion is to make it real simple to set it to save the stacked image to the same folder the images came from, and to name it with the file name of the first image in the stack followed by "stack of xxx images". I have attached my stacked result.  It is not perfect, perhaps due to movement causing blurry areas.

I tested it against Zerene Stacker using a stack of 95 images.  Zerene did better.

Thanks for the feedback!

It seems at least the second image has an alignment issue, any details on what option you used ? Was it a re-render with just the render mode changed ?
If yes, there was a fix around reusing alignment in the 1.2.0 that should be out today or tomorrow (currently going through validation), as well as improvements to the depthmap algorithm to better preserve details in low contrast areas (but that's obviously not the issue you encountered).
Would it be possible to give me access to a set of the original shots for testing ? (would be kept private or under NDA of course)

What would this bring to the table to differentiate it from Helicon and/or Zerene?

I am hoping to match them as a first step, and later on hopefully find a way to merge the wavelet and depth map algorithms.
The advantage of the wavelet algorithm is that there is no blur in the depth transitions. Part of that extra detail is taken from high-frequency information that comes through the blur/halos of what's in front, so it's not entirely made up interpolation or AI hallucination.
But whenever there are sharp transitions, the wavelet algorithms has a lot of ugly artefacts that extend way beyond the transition... I'm hoping to find a way to take advantage of both approaches, and ideally a way that is as automated as possible.

Betarover wrote:

I tested it against Zerene Stacker using a stack of 95 images. Zerene did better.

Thanks for the feedback!

Just like with Jim's images, it seems I have an alignment bug lurking in there... :/
Some of the softness should be improved in the v1.2.0 that should be out today or tomorrow (currently going through validation), but I am unsure where the alignment bug is coming from.

Would it be possible to give me access to a set of the original shots for testing ? (would be kept private or under NDA of course)

dhhddre wrote:

What would this bring to the table to differentiate it from Helicon and/or Zerene?

I am hoping to match them as a first step, and later on hopefully find a way to merge the wavelet and depth map algorithms. [...] I'm hoping to find a way to take advantage of both approaches, and ideally a way that is as automated as possible.

On that note, I'll be PMing a link to a stack that broke focusfusion once the files (11GB) have uploaded.

Let me know when you've retrieved them and I'll delete.

Message:

Yes you can have the files, whats best to send them to you?

Any way you prefer, and I can share a google drive folder with you in a PM if that's ok?

Alignment is my biggest gripe with tools, not when the camera does the stack, but handheld on eg bugs where movement between shots is high.

I have somewhere in the todo an advanced alignment for handheld, that could compensate for more camera movement, though it won't help if the subject moved (breeze, insect movement)

i also noted yesterday that your tool stacks regardless, canon dpp tries then says its failed to align

Right now it only gives up if it fails too bad.
I have started experimenting with ordering and discarding shots rather than requiring a nicely ordered sequence of shots: the goal would be to still be to stack when the focus point is moved by just moving the camera back and forth handheld (I'm not much fan of carrying a full size tripod, and mini-tripods only work for ground-level). You could then burst like 100 shots, the soft would then try to pick the top 10 or 20 that are most usable.

Thanks for the stacks! I have fixed the most glaring alignment issues on them, I will now work on the other issues.

odd as it may sound, when i'm handholding and manually merging, i do scan through the results and select ones of specific interest, eg right eye, left eye, feelers, back, tail, etc, so having automation for that and then doing the stack would be impressive

I was not thinking of something as sophisticated early on, but discarding the shots that don't align well or that have some motion blur, so that only the cleanest shots enter the stack (a bit like what some astrophotography stackers do).