I take quite a few snaps of pretty flowers and such, and while a swirly defocussed background is usually attractive, sometimes the defocussed front and rear parts of the flower are not (smile). So I've run some garage experiments with an automatically step-focussed and merged stack in the E-M1 using the Oly 60 macro lens at f2.8, and compared the results with a Photoshop align and blend of the same stack. Snaps are all JPGs, I've never been able to get any RAWs to look as good as the Oly ooc JPGs and have given up trying...
General setup
The above snap shows the general setup, I more or less filled the frame with the Lens Align target tilted at 20 degrees. The smallest steps on the target, from 32 up to 32 down, covered a depth of around 19 or 20 cm, meaning each step had a depth of around 3 mm. The "0" of the target was around 62 cm from the sensor plane.
Base snap, focus on smallest "0", f2.8, ISO 200
100% crop of base snap
I thought I focused on the "0", and the above crops show the depth of focus of a snap at f2.8 around 45 cm distant. In fact we can see that the focus point was just slightly above the "0", probably more on the smallest "1", and, depending on your definition of out of focus, we might think that the smallest "0", "1", and "2" are in focus, suggesting a depth of field of around 9 mm. Interestingly, the Android app "HyperFocal Pro" tells us that it thinks the depth of field is around 8 mm for this setup.
Oly in-camera 8-frame "differential = 2" stack merge
Photoshop 8-frame diff = 2 stack blend
I set the E-M1 to focus stack with a "focus differential" of "2", and snapped away. The resulting cropped in-camera merge is shown above, along with the Photoshop merge. In PS, I first aligned the stack using "Collage" as the projection (alignment) method, and then blended the stack using "seamless tones".
The Oly in-camera merge gives us an image with a depth of focus from around 9 or 10 up to perhaps 3 down, a depth of field of around 40 mm. The PS merge gives a slighter wider DOF, perhaps from around 11 up to 4 down. Both merges are reasonably pleasing, but interestingly, the align and merge algorithms of Olympus in-camera and Photoshop take quite different approaches to artefacts.
100% crop of Oly in-camera merge
Pixel-peeping the Oly merge shows strong halos, ghosting, or ringing at high contrast edges. We can see the ghosting around every digit, and inside most of the rectangles.
100% crop of Photoshop blend
Pixel-peeping the PS merge shows ghosting under almost perfect control, but instead shows a myriad of alignment and blending artefacts. We can see the smallest 5 up has a scrap of oof image tacked onto the right side of the "5" loop, and the rectangle holding the smallest 5 up has a "tab" attached to its top right edge. Similar artefacts can be seen elsewhere, such as on the top of the medium "0" loop, the rectangle edges between the medium 6 up and 4 up digits, and so on.
Oly merge "differential = 4"
PS blend diff=4
The results of shooting a stack at differential = 4 show us that the depth of focus has improved for the Oly in-camera merge to around 8 up and 8 down, but the PS blend gives us from around 22 up to 6 down. The Oly merge has failed somewhat in extracting the in-focus parts of the scene that the PS blend has successfully captured. Interestingly, the Oly merge shows something of an oof band running through 11 to 16 up, then a more in focus band from maybe 17 to 22 up -- which is where PS got to, but without the oof band...
Oly merge differential = 6
PS blend diff=6
The results of differential = 6 are along similar lines, the Oly merge pretty much failing to do its job, while the PS blend is far more successful, although we can now see oof banding along the target, at around 8 and 9 up, 14 and 15 up, and 21 up, and 7 down. I shot differentials of 8 and 10, but they added nothing to the story that we can see for differential = 6. I ran a similar series of tests with the Oly 40-150 shooting at 150 mm and f2.8, with very similar results.
Oly focus stacking with 40-150 Pro at 150/2.8
There, I used Photoshop's "Auto" for alignment, and unchecked "seamless tones" for the merge. I think the PS results this time around were better with "Collage" for alignment and "seamless tones" in the blend.
A clear result was that the Oly focusing algorithm distributes its 8 frames to improve focus in the front third and the rear two thirds of the scene relative to the nominal focus point.
What would I take away? (1) Provided the stack steps ("differentials") give good in-focus overlap to the series of frames, both Oly and PS give pleasing merges. The difference is in their artefacts. (2) Oly gives superior stack alignment so there are few and not so "funny" oof bits where you would not expect to see them. On the other hand, Oly ghosting on high contrast edges is dreadful. Photoshop is the opposite, terrific ghosting control, but oof "funny" bits peppered all over the place. (3) As the stack steps increase and so good overlap reduces or disappears, the Oly merge fails to give a satisfactory result, while the PS blend does the best it can, pretty successfully given what it is working with, with modest oof banding across the image where the lack of overlap was worst.
An issue remains -- what is meant by "focus differential" in Oly-speak? For my use, I've decided that I can set a focus differential to the f-stop I want to use. If I'm shooting at f1.4, differential = 1. For f2 or f2.4, differential = 2. For f2.8, differential = 3. For f4, differential = 4. You get it, up to f8 and differential = 8. I seem to get the focus overlap that the Oly in-camera merge algorithms need, and don't have to do the Photoshop thing.
For my artistic work, the Oly ghosting adds to the artistry! For my engineering work, though, it has to be a PS merge. It may be worth adding that, in PS, the mask set for each frame just before the blend can be adjusted in its layer. If you are willing to put in such detail work, you can eliminate all the PS "funny" oof bits and get a near-perfect ghost-free blend.
Unsharp Mask of earlier 100% Oly crop of in-camera merge with diff=2, amount 100% radius 10 threshold 0
If we don't want a layer by layer manual mask inspection and adjustment but do want to try and fix the Oly ghosting, we could try the unsharp mask in PS. The above image shows the result of a fairly gross amount 100% of a radius 10 pixel level 0 threshold Unsharp Mask on the earlier Oly merge. Ghosting almost eliminated at the cost of unrealistically crisp and contrasty edges.
