support test for Focus Stack
I continue to thave very bad results with focus stack...
Herebelow a simple 3 focus merge… do you have similar results?
I loaded the 3 source images on Dropbox (see below link). If someone can make a test with PR and/or other similar tools using the same images, we can try to discover if it is an issue related to my computer, to the source images or to PR…..
Thank you for any test result you might provide
I think there may be multiple things to consider
*[info] There is a gaussian roll-off in detail (sharpness) defined by the DOF. A shallower DOF will have a sharper slope, thus making it more difficult for the edges to be detected. (I am guessing On1 uses some sort of edge detection to determine focus in each image.)
* [user_process] At f/4.5 there is not much DOF; My guess, not knowing the scene, is three images in total is not enough. It would seem like each flower may need about 3-4 focus slices. In this case, the entire data set is only three. I would recommend 3 images for each flower, plus a couple between the second and third flower to get the branch detail
* [user_process] Alternatively, increasing to an f/8 or f/10 may extend the DOF enough to get closer. (I am guessing a tripod is part of the setup and the subject is motionless)
* [user process] It also looks like the brightness between each exposure isn't well matched... maybe it is the vignetting of the lens changing between exposures, I am not sure. In any case, the uneven background tonality between images is what is causing the random blotches in the zones away from the flowers. (Technically, I would put this as an On1 algorithm issue, but there are also ways to reduce/avoid this situation using the camera hardware and/or technique
* [On1 issue] In the FWIW column, I have decided on other tools for "precision" focus stacking (i.e. close up work like bugs, flowers, etc.). I have used on1 for "landscape focus stacking" with OK results, but don't trust it as the precision focus stacking seems to be a bit less developed than competitive tools.
* [On1 issue] I have pushed a 54-shot image cover about 2cm of distance into On1. Of course, I had to reduce the images to less than 10MB else On1 would choke/die (other tools I used 100M x 54 files without issue). Even with this level of detail, my experience is On1 is still developing their algorithms.
If On1 reads this, it would seem the "between layer interpolation" algorithm needs a bit of work. Applying a bit more math (specifically, calculus and matrix algebra) and experimental modeling using MatLab, SciLab, MathCAD, other could go a long way into developing a >photography specific< matrix interpolation algorithm.0
Thank you Gus for the explanations.
I tried to increase DOF to f/8 and feed ON1 with 10 pictures instead of 3.
Herebelow the result….unfortunately no improvements appear.
I also tried using Photoshop (see second image) and the results are much better (some adjustments are still needed in the background but are manageble in post production or using more images)
With PR 2019.2
As a test, it may be worth trying the 3-image stack in PS, too. I bet PS may be happier with the 10-shots at f/8 as compared to the 3-shots at f/4.5. This step would confirm if it is worth the time to do the extra shots. In my experience, it is worth it, but your mileage may vary.
Your On1 experience is exactly like mine. The On1 algorithm is a few steps behind. In this case, it seems even worse. The blotches are obvious. The flat colors on the stem and leaves show even worse. The lack of dynamic range in the flowers is horrible and the contrasts are all all off.
In comparison, the PS image hiccup seems most obvious on the nearest flower. But indeed, significantly closer to a workable image.
I find the similar issues with the On1 HDR (which also stacks images). The panorama feature is less robust, too. I am always worried about confirmation bias, but this enforces thoughts of layer to layer number crunching of On1 still needs help.0
I’m not a fan of focus stack or hdr in ON1 - horses for courses: Affinity Photo for focus stack and Aurora for hdr.0
I ran 2 tests see results on your dropbox:
PR 2019.2 - results similar to yours - file name pfeiffep_fs test.onphoto
PSE - great results - file name pfeiffep_varie_PSE.psd
I'm waiting for the 2019.3 to test focus stacking since 2019.2 results are awful0
herebelow the three images stacked with Photoshop. The result is quite good in therm of focus stacking even with just 3 shots f / 4.5 (do not mind the colors since no editing has been performed before and after the test)0
Peter I can not find your tests in dropbox...
Let's hope PR.3 will be better in both focus stacking and HDR (at the moment I find more powerful the "HDR look" filter in the Effect section)0
I think I still like the 10-image stack in PS better... But the 3-mage stack in PS is respectable.
Your tests definitely support the need for stacking improvement (hopefully) in PR.3. Thanks for sharing the results.0
I don't know what happened to the posts.
I ran 2 tests just to prove to myself that there was nothing amis with your dng's. The results obtained with Photoshop Elements with Elements+ add on were perfectly acceptable and probably about equal to your results using full blown Photoshop.
Just an aside - I've yet to have an acceptable focus stack using PR.
This is my latest FS - 20 images stacked with PSE0
Thank you Peter, great picture!.
I agree with Gus, the higher the number of images the better the result is.
I think that the selection of a certain DOF should depend to way you want the background appear.
For sure you make the work easier to the software using higher FOW (f/10-f/16) but, if you prefer to have a blur effect on background, you can compensate using more images taken at lower DOF0
Yes I'm waiting for the update as well.
I'm finding a weird situation when I open the folder with the stackable images....click 'select all' and all the images are selected with the exception of the first image which had a yellow outline.The stacking procedure ignores ths first image [which from a focus point of view is the most important.]
I'm getting much better results using Picolay using default settings0
Please sign in to leave a comment.