Synthetic stereo in Zerene Stacker

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Synthetic stereo in Zerene Stacker

Post by rjlittlefield »

Consider the following two sets of crossed-eye stereo pairs:

Pair #1:
Image

Pair #2:
Image

The subject here is a paper wasp mandible, total frame height 1.8 mm.

If you're like me, you probably have trouble telling the difference between the two pairs. The first one has a little more halo around the edges of the mandible. It also has a little less "shimmer" in the center of the mandible where those big reflections are, and around the ball joint at the base, and especially on that OOF light stuff at bottom right. Pretty minor differences, visually.

The exciting part comes in how each pair was produced. Pair #2 is traditional true stereo, two stacks shot about 6 degrees apart. But pair #1 is completely synthetic, made computationally by Zerene Stacker from a single stack.

The button-pushing of how to do this is described at http://zerenesystems.com/stacker/docs/S ... Stereo.php.

Very briefly, you tell ZS to run the stack two or more times, introducing a bit of "shift" each time to make the composite image look as if it were shot from slightly different viewpoints. Then you take the resulting composites and make a stereo pair, or a movie, or maybe even a stereo movie.

Synthetic stereo has several important advantages over the traditional scheme of shooting two actual stacks. These include:
  • You get stereo almost for free -- no additional time to setup or shoot, and no additional space for image files.
  • You can even reprocess old stacks that you shot with no intention of ever doing stereo.
  • You can get stereo even when shooting through an ordinary microscope that only provides one view.
  • The left and right images are guaranteed to correspond -- no problems with different stack depths, different magnifications, different stuff in the background, and so on.
  • You can orient the subject however it's convenient to illuminate and shoot, then display it however it looks best. (With the traditional scheme, you can't change the orientation after you do the shoot.)
The method is not foolproof, of course, but it's surprisingly effective, particularly when used with the PMax stacking method to give robust handling of overlaps and bristles.

It also has the quiet charm of working best when you need it most, with high magnification deep stacks that have shallow DOF per frame.

All images shown here were done with Zerene Stacker PMax.

The two pairs shown above are full frame and are unretouched except for framing and sharpening.

The following two pairs are cropped less than full frame, using StereoPhoto Maker, but also are unretouched except for framing and sharpening. Both are "from the archives", originally shot with no intention of producing stereo.

My friend the Hobo Spider. Field width shown here is 6 mm.
Image

Some "Tinier spring flowers", field width 2.5 mm. (Yes, these are very small flowers!)
Image

The flower example is especially interesting because the background looks "flat". That's because all of the background pixels essentially come from one frame -- the farthest one back -- so they all have the same disparity and thus the same apparent depth. If you're shooting a new stack with intention of rendering in stereo, and you have some texture in the background, you might want to shoot a few extra frames behind the subject to get some separation between subject and background in the stereo rendering.

A couple of other recent examples that I posted out without explaining include the "Spectacular spider-killer-killing wasp" and the earlier paper wasp mandible shown HERE.

I hope you find this interesting -- it was fun preparing them. :D

--Rik

Photography details:
  • The paper wasp mandible was shot with Canon 300D camera using a Nikon CF N Plan Achro 10X NA 0.30 objective on 200 mm extension, 0.00033" focus step, stacks of 101 to 118 frames. Illumination was with dual 60-watt desk lamps through a Kleenex tissue diffuser.
  • The hobo spider face was with Canon 300D camera using Olympus 80 mm bellows macro lens with Olympus 170mm matched auxiliary lens, marked f/5.6. Halogen dual fiber illuminator with kleenex tissue diffuser. 61 frames spaced at 0.005".
  • The spring flowers were shot with Canon 300D camera, Olympus 38mm bellows macro lens at f/5.6, 39 frames at 0.002" focus step.

Edit: to add size info for the hobo spider and flowers.
Last edited by rjlittlefield on Mon Aug 17, 2009 12:30 pm, edited 1 time in total.

g4lab
Posts: 1437
Joined: Fri May 23, 2008 11:07 am

Post by g4lab »

I always wondered if it was possible to do this. Very interesting and exciting to see that it can be done now with relative ease.

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Here is another example that I shot today specifically as an example of mineralogical use.

The subject is a chunk of local lava, the frothy stuff that landscapers use for groundcover.

Full frame, 17 mm field width:
Image

Crop of the right central portion (using StereoPhoto Maker), field width about 8 mm:
Image

Stack length is 79 frames with a focus step of 0.005 inches, using an EL Nikkor 50 mm f/2.8 at f/4. Stereo at +-5% maximum shifts.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

g4lab wrote:I always wondered if it was possible to do this. Very interesting and exciting to see that it can be done now with relative ease.
The basic techniques go back to the 1980's. That's when NASA developed the pyramid algorithms, and when I started publishing about stereo and motion in data visualization.

What you're seeing now is primarily a matter of algorithm selection and packaging.

But yes, being able to do it easily makes quite a difference.

Here is a stereo rendition of the automobile headlamp that you enjoyed.

Image

This was rendered at +-1% shift using DMap, unretouched (but with a carefully tuned contrast threshold to get a clean background).

--Rik

g4lab
Posts: 1437
Joined: Fri May 23, 2008 11:07 am

Post by g4lab »

The coiling on the headlamp becomes much more apparent in the stereo view.
Rik do you know any reason why I can always fuse crosseyed stereo images but not always straight ones??

G

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

g4lab wrote:do you know any reason why I can always fuse crosseyed stereo images but not always straight ones??
I have the same problem.

For me, it's most often a matter that the image got reproduced with a separation wider than my eyes, so I have to diverge, not just look straight ahead. This is very common with online images due to differences in monitor resolution. Somebody prepares an image on a monitor with small pixels and it works fine for them, but nobody including the original author can fuse it up on a regular desktop monitor. Or maybe the author happens to have eyes that are set a bit wider than mine, so that even on the same monitor they could see it and I couldn't.

The other issue is what I might call "error margins". With crossed-eye viewing, there's a huge margin for error because I can converge about twice as far as required by most pairs. But with straight-ahead viewing, there's almost zero margin because I can diverge hardly at all.

--Rik

Joseph S. Wisniewski
Posts: 128
Joined: Fri Aug 15, 2008 1:53 pm
Location: Detroit, Michigan

Post by Joseph S. Wisniewski »

A normal human can track about 30 degrees inward (cross eyed). Cross eyed viewing exagerrates "keystoning" in stereo images, and takes extra work to compensate for.

To track outward is much harder. That's called going "wall eyed" (yes, it's named after the fish). We're programmed not to do that, because tracking past infinity can just leave the eyes wandering aimlessly and uncoordinated.

"straight viewed" images can only be viewed unaided (by most people) if the images are spaced about at the interocular distance. About a 3.5 inch wide card is the limit for most people, or 4 inches on the screen. (There's a reason: the screen becomes "infinity" easier than a card). To view those, just bring the card so close to your eyes (or your eyes so close to the screen) that the image gets totally blurred, then pull back, it will snap right in. That technique is used for printing stereo cards at "business card" size, and I use it for "freebie" stereo images.

Periscope viewers have (obviously) fixed or adjustable periscope mirror arrangements to get the centerlines of your eyes farther apart and sometimes to also give you the ability to go "wall eyed". My favorite is a little "Wheatstone" viewer with an adjustable double periscope mechanism that lets you go from straight ahead to about 20 degrees wall eye. 20 degrees is enough to fuse stereo on full screen images on a 30 inch LCD. Periscope viewers are relatively heavy and adjustable ones like the Wheatstone have a fairly narrow field of view, but they give you a very clean view with no chromatic aberration or distortion.

Prism glasses give you a substantial "wall eye" view. I have a pair I got from Berezin 3D that is about 20 degrees, enough to fuse full screen stereo pairs on my 30 inch Dell. They have a wide field of view, and are quite light, but have annoying chromatic aberration. Achromat prism viewers are possible, but I've never actually seen them marketed...

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Today I was revisiting these cricket mandibles for other reasons, so I decided to make a stereo pair from the stack.

Image

I see that in the original rendition, I decided to leave the farther back parts OOF to avoid confusion. No need to do that here because the stereo provides the separation, so I just ran the whole stack.

This is at +-2%.

--Rik

Joseph S. Wisniewski
Posts: 128
Joined: Fri Aug 15, 2008 1:53 pm
Location: Detroit, Michigan

Post by Joseph S. Wisniewski »

Rik, three comments...

1) I love the lamp filament. That's one of the best examples of notching I've ever seen. Back in the 80s, was experimenting with anti-notch automotive lamp drivers back at Microdot Elicon (long before they got bought by Tyco).

2) You mention StereoPhoto Maker. (Oddly enough, it's what I used to view all the stereo pairs in this thread). If you actually like the program, please write Masuji SUTO and tell him so. Although he seems to be happier (or at least more productive) theses days, I think he's still upset by the whole experience with the "cease and desist" letter. (I think it was from Real3D, but it might have been from the SIFT folks at UBC).

3) On the first image, I like the synthetic 3D version better, mostly because the reflections follow the surface. I know that's not natural, but it makes it so much easier to fuse than the "real" 3D version. Stereos of highly reflective "Escheresque" objects tend to fail...

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Joseph, by the numbers...

1) Thanks! This was my first close look at a halogen filament. I was surprised, to say the least, at how complex and interesting the shapes are.

2) Done. Thanks for the suggestion.

3) I agree. I've also tried a couple of examples on people who don't spend much time looking at stereo pairs. They couldn't say what the difference was, but their immediate reaction was that "pair A" (the synthetic) was definitely easier to look at.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Today while preparing another posting I happened to revisit this cherry fruit fly ovipositor.

That prompted the memory that this beast has some anatomy that might be made more clear in stereo.

So I ran the original stack through the newest Zerene Stacker and worked up a pair.

Image

What I had hoped to see in stereo was a better indication of the longitudinal groove in the dorsal side of the ovipositor (right side, as shown here).

That aspect shows up very nicely.

But I was surprised at how clearly the stereo pair shows off the transparency and concavity of that membrane on the thicker section. I already knew the membrane was transparent because I could see background colors through it while preparing the whole-body shot (image #2 at the earlier link). But I didn't realize it was concave or that the transparency would be made apparent by the way the membrane refracts the brown structure behind it. I guess I'll know next time to look for that sort of thing.

This is an asymmetric pair, -1% and +3%. I needed both views to clearly show the back side of the grove on the ovipositor, which limited the shift to -1% in one direction. But then I wanted more depth, and pushing the other direction to +3% worked out well.

Cropped in StereoPhoto Maker, sharpened and framed in Photoshop, no other retouching.

Total field height in these images is about 1 mm. The earlier single shot has an accurate scale bar.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

This is a new pair, generated specifically to serve as an example for mineralogical use.

Of course this isn't really a mineralogical sample, it's actually a small piece of sculpture apparently made by pouring molten brass across a bed of assorted small rocks. Oh well, close enough! :lol:

Not quite full frame, this crop is 21 mm wide:
Image

A closer crop, about 7.5 mm wide:
Image

The source stack was 130 frames with a focus step of 0.010 inches, so total depth = 1.3 inches = 33 mm. Shot with Canon 300D at a magnification of 0.8X using an Olympus 80 mm bellows macro lens at f/4. The stereo pair is at +-3%, PMax, no retouching. Stereo cropping in StereoPhoto Maker, framing in Photoshop.

Here is the on-axis view (+0%), again not quite full frame, PMax, no retouching.

Image

I hope this is helpful.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

Here's another old stack that I had occasion to reprocess today.

Image

This stack was originally shot at f/8 and about 1.2X, 49 frames. The images shown here are slightly cropped, to about 12 mm frame height.

The pair is slightly asymmetric, -3% to the left, +2% to the right. The result at +3% to the right had too many artifacts even for presentation at this size.

The above pair was rendered with the DMap method, no retouching. PMax did not work well on this subject. It gave too much contrast enhancement and introduced broad light halos around the dark seeds.

The original thread, which includes a couple of true stereo pairs, can be found HERE.

--Rik

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

I had occasion today to reprocess another old stack, this time the papillate stigma of a maple flower as originally posted HERE.

Image

To save time while making the stereo pair, I set Image Pre-sizing to 33.33% and loaded only every third source file. The resulting stack of only 27 small frames processed pretty quickly, about 5 minutes for 9 outputs from -4% to +4%.

This is the pair at -4% and 4%. Slight crop, cloned out a couple of dust trails, and cloned out a couple of stereo anomalies along the outer margin.

I was more surprised than usual to see how flat the single images look, after seeing the stereo pair. Apparently this is one of those subjects for which most of the non-stereo cues about depth and structure don't work very well.

--Rik

Tesselator
Posts: 388
Joined: Sat Mar 27, 2010 11:40 pm
Location: Japan
Contact:

Post by Tesselator »

Both of those last two seem to stick out off the monitor. Mostly (since your butterfly taught me how to view these) they seem to sit behind where my brain thinks the monitor is. Neat!

Post Reply Previous topicNext topic