Cross-eyed Auliscus sculptus and Navicula

[WojTek]

Hello James,
I prefer the red/cyan and just love your 3D photos!
Best, ADi

[Smokedaddy]

Hello James,
I prefer the red/cyan and just love your 3D photos!

Thanks WojTek,

Lot's of luck on my end. I thought the Auliscus turned out nicely as an Anaglyph red/cyan too. Here is the Navicula in a grey red/cyan Anaglyph and a color Anaglyph red/cyan.

-JW:

[Pau]

I still have problems with your anaglyphs, reversing the red/cyan glasses i see them much better

[Smokedaddy]

I still have problems with your anaglyphs, reversing the red/cyan glasses i see them much better

Just curious, does this one make a difference? I swapped the L/R image before making the Anaglyphs. The only difference I see between the two, being one has a recessed 'frame" and the other protrudes. I also don't believe the Navicula is a poorer 3D image compared to the Auliscus.

[Pau]

Thanks for the new image!
It seems that this one has too much displacement, so the 3D effect is not better than with the glasses reversed. No problem, just I was curious.

[Smokedaddy]

Thanks for the new image!
It seems that this one has too much displacement, so the 3D effect is not better than with the glasses reversed. No problem, just I was curious.

I'll have to investigate this. The Auliscus red/cyan Anaglyph looks outstanding to me but the Navicula Anaglyph FRAMES are either popping out or in IF I flip them in SPM. I'll have to ask one of the experts I know and see what the issue is.

[iconoclastica]

Once in a while I have problems exporting the smaller sized versions that can be uploaded here. The first time I don't remember quite well - think it had to do with the jpeg quality. Second time was a short while ago and particularly stubborn. It was solved only when I changed the colour space from Adobe to proPhoto...

[Smokedaddy]

Sorry for all the photo clutter. Here's yet another Anaglyph test. Notice the FRAMES on these are not popped out or popped inward.

[Pau]

I ignore what you did, but the last anaglyphs work clearly better for me

[rjlittlefield]

So, is everybody happy with the last anaglyph, THIS ONE ? I am too, FWIW.

And do everybody's glasses have red lens on the left?

If yes to both, then I would like to talk about some details.

--Rik

[iconoclastica]

So, is everybody happy with the last anaglyph, THIS ONE ? I am too, FWIW.

And do everybody's glasses have red lens on the left?

If yes to both, then I would like to talk about some details.

--Rik

With regard to the making of the anaglyph version: yes. Starting with the topmost stereo pair, I get almost the same result, wearing the spectacles I daily use.

[Smokedaddy]

So, is everybody happy with the last anaglyph, THIS ONE ? I am too, FWIW.

And do everybody's glasses have red lens on the left?

If yes to both, then I would like to talk about some details.

--Rik

With my cheap glasses, red on the left, cyan on the right, the Diatoms all look similar. Meaning the Diatom itself. In the other Anaglyph's I posted, meaning the last group, the only difference I see is the FRAME (or border). To me, all of the Diatoms look the same, but the frame/border is either protruding out of the screen in the foreground, with the Diatom in the background, OR, the opposite case, frame/border protrudes into the screen with the Diatom in the foreground. When I view all of the Navicula Anaglyph's, there isn't much relief in any of them, so they look the same to me, unless I haven't studied them closely enough. I'm out of town for a while, so I won't be visiting the site more than likely.

[Smokedaddy]

I should mention that by no means am I an expert generating anaglyph's. Just clicking buttons in SPM. I have created a LOT of anaglyph's unrelated to microscopy that have turned out excellent using StereoPhoto Maker, so I assumed I was doing everything correctly. Those are Cherry Trees just in case you're wondering.

-JW:

[rjlittlefield]

So, is everybody happy with the last anaglyph, THIS ONE ? I am too, FWIW.

And do everybody's glasses have red lens on the left?

If yes to both, then I would like to talk about some details.

I see only two replies to my questions, so I assume not many people are eagerly following this thread. But proceeding anyway...

JW wrote:

I also don't believe the Navicula is a poorer 3D image compared to the Auliscus.

I disagree, and I think I can illustrate productively why that is. To do that, I'm going to extract the left/right views from stereo format, and animate them as rocking displays.

To begin, let's look again at the Auliscus pair. Here it is, rocking. (This is animated GIF, at 3 seconds per view. If it's not rocking, check your browser settings.)



For me, there are not many surprises in this view. It conveys a strong impression of a 3D subject that is switching between left-eye and right-eye views, just as if I were looking at a real thing and alternately blocking one eye, then the other. The only notable exception is one small area at extreme right, where the shape of the projecting nubbin changes oddly between the two views.

For comparison, let me reach into the technical archives and give the same treatment to a true stereo pair. (This is from viewtopic.php?t=33871, the 4 degree case of a small ball covered in grease and dried spices.)



Again, no surprises. This just shows a physical rotation of a real 3D object.

But now let's look at the Navicula pair:



In contrast to the other two, the Navicula pair does not give me any immediate perception of 3D rotation. Instead of toggling between what clearly appear to be left eye and right eye views of the same subject, what I see is that it toggles between two images that are just different from each other. Dots in one change to streaks in the other, the subject appears to lean slightly, and parts near the upper tip change shape in ways that I have trouble making sense of. There is even some vertical disparity between the two views -- features move up and down in the frame, instead of just moving laterally as they would do in clean stereo.

I think that all these issues combine to produce the ambiguity that appears in the earlier discussion.

Referring to his switched-sides anaglyphs, JW writes that "To me, all of the Diatoms look the same, but the frame/border is either protruding out of the screen in the foreground, with the Diatom in the background, OR, the opposite case, frame/border protrudes into the screen with the Diatom in the foreground."

This is a telling observation, because inverting the depth of the frame also inverts the depth of the diatom as indicated by stereo disparity. If the stereo cues were strong and unambiguous, then switching sides on the Navicula would change the diatom's appearance between convex and concave, at the same time that the frame/border pops forward and backward.

But in fact the stereo cues are weak and conflicting. Meanwhile, both prior knowledge and shape-from-shading quietly but strongly suggest "nope, that diatom is convex both ways". The result is that in this case stereo disparity conveys little or no information about 3D shape, other than to indicate that there must be some because what else could be causing differences between the images.

I hope this helps to see what has gone wrong with the Navicula.

Maybe it also illustrates a useful technique for checking the quality of stereo: if switching sides shows effects other than the subject rotating, then there will be problems with stereo also.

Stepping back from what, it may be helpful to wonder why the Auliscus came out so well while the Navicula did not.

I think the main reason is that Zerene Stacker's synthetic stereo method has not worked well for the Navicula. In general, synthetic stereo is vulnerable to stacking artifacts when there are structures that have foreground/background overlap or deep pits. Those artifacts may not be troublesome when they are isolated, but with repeating patterns like the pits of this Navicula, the artifacts can accumulate in strange ways. I think that's what produced dots in one view versus streaks in the other view.

As a secondary reason, I think the Navicula pair suffered some corruption at the hands of StereoPhoto Maker. JW mentioned to me offline that "I'm positive I realigned it in SPM", and that raises a number of warning flags. The alignment process in StereoPhoto Maker is designed to fix stereo defects such as keystoning, camera roll, and vertical disparities that are common with true stereo. The way it does that is by finding matching features in the two views and calculating what transformation is necessary to remove the defects. That's great if such defects are present and SPM correctly figures out how to remove them. But with Zerene Stacker's synthetic stereo, defects like that never appear in the first place, while instead there can be other defects that lead to SPM making incorrect decisions that end up introducing keystoning, roll, and vertical disparities, instead of removing them. My best guess is that SPM looked at the Navicula pair, got confused by the stacking artifacts, and transformed the images in ways that it erroneously thought was an improvement.

Best advice is that when working with Zerene Stacker's synthetic stereo, do not ask SPM to make any adjustments other than left/right shift to change overall depth position with respect to the display surface. I have added this caution as a note at https://zerenesystems.com/cms/stacker/...#tips_for_best_results .

I hope this helps!

--Rik

[iconoclastica]

If I use your formula to calculate the view angle, I find 0.95% either way, inserting average stack depth of 25µm and framewidth = 183.5µm. Should I crop the image before generating the stereo pair, the frame width would become one third of the original and so the view angle about 3%. Experimentally, I found 1% to be optimal, so the original frame width seems the measure to use, or rather, the cropped frame should have used a smaller view angle. How does this angle relate to the frame size?