Fun with Focus Stacking

[PAL_Houston]

Going to higher f-stops and longer exposures wasn't giving me enough depth of field, so last night I broke down and started playing with focus stacking.  I experimented with different numbers of stacked photos, and little bit with some of the processing options provided by CombineZP software.  This is the best I've done so far.  It is a stack of 8 photos of Oregon on my C&I Sub, using f5.6 aperture and manual focusing with a Sony A-33 and 18-55mm lens.  The stacking options were "align and balance (thorough)" + "Pyramid Maximum Contrast", but being a novice to this I don't really know what that means.



What have you done with focus stacking, and how do you get your best results?

[wm3798]

DKS did this one for me, which appeared (obviously) on the cover of N Scale a couple of years ago.

Lee

[DKS]

Although I'm not an expert by any means, I have tinkered with Helicon enough to be aware of the pitfalls of image stacking. Among other things, I've learned that it's better to take more images at a larger aperture than fewer at a smaller one, as the latter will introduce a degree of softness. Also, it's best to avoid hard-edged objects in the foreground, as they can produce halo artifacts. While the following images are not compositionally the most interesting, they did serve as good learning exercises. They were all shot with a 50mm lens at f/22.

This example was made with 6 images:



The foreground objects are about three inches from the lens, while the scenery in the background is about two feet away.

This one comprises 11 images; some halo artifacts are evident around the sign:



This is a stack of 10:



For the following image, I deliberately allowed the background to fall out of focus to add more interest, and only took 4 images:



I applied what I'd learned to the shot I did for Lee, which involved 12 images taken at f/8:

[wcfn100]

The best one I've done might be this one (please excuse the extra effects).



(click for super-size)

The reason I think it worked well is because the main focus is mostly on center and the depth isn't too exaggerated.  I think these things could be said about Lee's shot as well.  This is an easy formula for success. 

When you start getting into situations where you have a main object in the front and strong to one side and another fairly deep and to the opposite side of the frame, you can start to get a lot of anomalies.  Not to say they can't be overcome, and every shot is different, but there are things like that that make it harder.


Jason

[jagged ben]

Anybody here tried using a Lytro to photo model trains? 

[DKS]

Rich Yourstone published an excellent article on image stacking in N Scale magazine some years back. Well worth hunting down.

[Mark W]

Anybody here tried using a Lytro to photo model trains?

I was this close to ordering a Lytro a few weeks ago.  However, using Helicon or Combine Z allows for the exact same results (when used properly). 
I think the only advantage the Lytro would provide is to eliminate the halo artifact described by David above.  Of course the Lytro would complete it all automatically, but you'd also lose the image quality and high resolution of a good DSLR. 
Maybe in several years when the technology improves 4 times over and the price comes down to 50 bucks. 

Here's my last attempt using Helicon.  Stacked with 21 photos with a long lens, making each photo have a relatively small depth of field.  I think that is the reason for most of the artifacts that can be found.  More experimentation will follow.

[peteski]

The halo effect is unavoidable because the focused part of the image needed to fill in those areas is not available in the stack of photos.  The only way to recreate that info would have to be done by the application itself.  Sounds like Lytro might do that.

In this example, the halo effect is visible around the corner posts of the bumper car arena.


In this shot I just took enough stack photos to get a focused view of the arena. I didn't bother taking focused shots of the background beyond the end of the arena (so the further background looks fuzzy).

[jagged ben]

Of course the Lytro would complete it all automatically, but you'd also lose the image quality and high resolution of a good DSLR. 

But maybe I don't have a good DSLR and probably won't get one?   

Also the point of the Lytro would be that you don't have to take multiple shots and could even take pictures of moving trains.  Or is that expecting too much?

Have to get back in touch with the friend who has one.

[jagged ben]

The halo effect is unavoidable because the focused part of the image needed to fill in those areas is not available in the stack of photos.  The only way to recreate that info would have to be done by the application itself.  Sounds like Lytro might do that.

I could be wrong, but I believe the Lytro actually takes light readings at all focal points and thus doesn't have to recreate anything. 

[rsn48]

For wonderful depth of field, I have found my Android phone, HTC One, takes great pics.  The lens is near the top of the phone, but by turning it upside down, its at the bottom.  Put the phone at or near track level and you have great depth of field, an incredibly small sensor and wide lens make for a very wide lens and high aperture settings.  With my phone, I really need to stabilize it as any motion will soften the image, so bright lights for a faster shutter speed.

[pnolan48]

I've been using Helicon for years, or since Rich Yourstone more or less introduced it.



Prior to Helicon, I'd been manually stacking (and adjusting) images, which is a real pain. I learned that sometimes I had to cut out some intermediate images (as well as the far and near images) to get the best results. A 24 stack of images might print best with something like this------5, 6, 7, 9, 10, 11, 14, 15, 18, 20. I learned this technique by trial and error on the railings of my ships, and other very repetitive items like fence rails. Helicon had (probably fixed now) a problem with rendering things like ship railings, which I certainly experienced firsthand! And the sequence, and what to cut, was pure trail and error

Other tips: Helicon works best with the sharpest f stop, usually either side of f/8. This requires more slices, and sometimes more work eliminating some intermediate slices. It's clear, when I watch Helicon work, that it is choosing the pixels that are "smallest", under the theory that they are the sharpest. Probably theoretically true, but with intricate images (like railings) this sometimes results in jouncy or overlapping lines, which can be corrected by eliminating slices. I understand why this happens, but can't spend the rest of the night trying to explain it.

Helicon's greatest impact is with sharp, medium-length lenses on a DSLR. With a 10 mm lens on a D90, I can get focus from about 3 inches to infinity at max f. With Helicon I can probably add an inch or two at the close focus. Sometimes it is worth it, as here, with the close-in bushes taken in slices at f/8:



But the image of this taken with the same lens at f/22 was fully acceptable.

Which brings me to my last point tonight: it's the glass that determines how any images comes out. I shoot nearly everything with my Nikon 10-20 zoom because none of my other lenses come close to it, Helicon or not.

[pnolan48]

Oh, I forgot this technique with Helicon, which I found considerably improved the image in the foreground:

I would shoot the close-in focus slices at f/8 (if that was sharpest), then the rest at f/11 or f/16. I always used Helicon with a fixed manual exposure (so the backdrop didn't go crazy), so switching between f/8 and f/whatever was just a manual adjustment (that I often forgot).

This technique seemed to make the foregrounds a little more crisp. Helicon does auto exposure adjust; I just found locking the exposure in created better results, especially since I was switching apertures.

[laconj]

I love Helicon!!

[skytop35]

Rich Yourstone published an excellent article on image stacking in N Scale magazine some years back. Well worth hunting down.

It was N Scale Railroading Jan/Feb 2006