Stacking with post focus on G80

[davidedric]

I think most of you will be familiar with this idea, but if not here's a brief explanation.

Some Panasonic cameras, in this case the G80 (or G85) have a "post focus" capability. The camera surveys the scene and establishes a series of focus points (I need to read up on how it decides what a significant focus point is), then it runs a two second 4K video, with each frame at a different focus point. You can process the video in camera, but I prefer to do it on the desk top.

I used Photoshop to render the video into 8mp stills, and decided by inspection which frames to include for stacking. I used Helicon focus for the stacking. There are two examples below. They are nothing especially remarkable, except that the first was shot from a tripod, and the second handheld.

They had some minimal processing in Lightroom, before exporting them. I'm sure that I could improve them, like the ghosting on teh tripod photo in particular, but I was amazed to get decent results handheld - it does I think become a practical tool. It may be that having a stabilised lens and IBIS helped.

(edit - just realised that there is no Exif - both shot at 35mm (70mm full frame) at f2.8)

Dave

[xpatUSA]

Thanks for posting, Dave, nice illustrative examples.

I've been looking at similar stuff lately and have found auto-alignment very useful as found for example in 'Hugin' a panorama/stacking app. Doesn't help if the ghosting is caused by significant scene local movement, though.

How many images did you end up stacking from the tripod shot, may I ask?

[davidedric]

Thanks, Ted

I used around 25 images. I think the ghosting is an artefact - there wasn't any local movement. The camera takes as many frames as it can find to focus, but then wanders off and just fixes somewhere, I think I might have included the first wanderer. Not sure - early days

I'm only just starting to use Helicon again. I tried one of their other stacking modes, and that seems to have eliminated most of the problem.

Dave

MOD COMMENT
I have removed the quoted images in Ted's post and added the original here for side by side comparison in LyteBox because this makes the haloing Dan discusses below far easier to identify for members not yet attuned to recognising stacking artefacts - best done in LyteBox, where both will be same size and you can switch between them with arrow keys.
Another Dave

[DanK]

Interesting. I think the effectiveness of the auto-alignment will be critical. All of the stacking software I have used has this function, but in the packages I have used, the amount of misalignment it can handle well is limited.

Haloing, such as that on the stem and around the edges of the outer petals, may not reflect movement. One can get stacking artifacts that look like that with no alignment issues at all. I think it is a function of how the stacking algorithm handles edges between areas at different distances. For example, in the software I now always use, Zerene, one of the stacking algorithms is more prone to halos of this sort than the other. Given that the first is superior in other respects for my work, I often composite using both and then paint from the second onto the first to deal with halos.

[xpatUSA]

Interesting. I think the effectiveness of the auto-alignment will be critical. All of the stacking software I have used has this function, but in the packages I have used, the amount of misalignment it can handle well is limited.

Currently I'm using the popular command-line "auto_align_images.exe" (used by many GUI apps) and in "verbose" mode it shows all the searched for "contact points" how many good, how many bad - and the final congruence(?) in pixels or fractions thereof.

[DanK]

Currently I'm using the popular command-line "auto_align_images.exe" (used by many GUI apps) and in "verbose" mode it shows all the searched for "contact points" how many good, how many bad - and the final congruence(?) in pixels or fractions thereof.

I believe my exposure fusion LR plugin uses that alignment tool. I haven't used it for focus stacking. The proof, however, is in the pudding: what the composite image looks like. And I have had serious problems with halos in the case of some images in which alignment was not an issue at all.

[Dave Humphries]

How many images did you end up stacking from the tripod shot, may I ask?

Good question Ted, I was wondering that too.

Another Dave

PS
Hope you don't mind, but I didn't see the benefit or need, in this particular instance, of quoting Dave's post, the additional pictures not helping when viewing in LyteBox and going between images.

[davidedric]

I assume you saw my answer.

I just try to choose the start and end images (viewed using FastStone) and then throw them over to Helicon. I am sure I can learn more, but processing 25 or so 16mp images takes no more than a minute or so.

Dave

[xpatUSA]

I assume you saw my answer.
Dave

Thanks, Dave (he of the black Stetson), 25 images, IIRC.

[Dave Humphries]

I assume you saw my answer.

I did too thanks.

I assume "a couple of seconds of 4K video" generates them at 25 or 30 fps?
So that's around twice as many as you actually use, presumably to save on processing?

[davidedric]

I did too thanks.

I assume "a couple of seconds of 4K video" generates them at 25 or 30 fps?
So that's around twice as many as you actually use, presumably to save on processing?

Yes, that's right - I think it is 30fps - but processing time isn't the reason to exclude some.

So fast I can tell, this is what happens. The camera identifies all the planes of focus it wants to use, and then runs the video starting with the nearest. When it runs out of planes it completes the video focussed on the furthest plane.

So more than half the frames were of a perfectly exposed brick wall. I excluded them because I wanted to keep the background blur, but if I'd wanted it sharp, I would probably have included just a few more.

Dave