Nezara viridula 10x, 12x, 19x and 40x

seta666 · Post by **seta666** » Mon Feb 14, 2011 4:24 pm

A green stink bug, I did a 10x but I found it interesting as it was covered in polem.
Nikon BD plan 20/0.40, 165 shots with 0.005mm steps. 3 ikea led lamps and EFCS

View Large
http://farm6.static.flickr.com/5176/544 ... 7b7f_o.jpg

And the stereo
http://img577.imageshack.us/img577/8394 ... xcopia.jpg

Also I did a 10x with mitutoyo 10/0.28, 127 shots with 0,01 steps. Same lighting. If you see some weird stuff is because I cloned some dust and I am not that good in PS. Last shot with aperture closed

View Large
http://farm5.static.flickr.com/4075/544 ... 8a3f_o.jpg

Here is the stereo
http://img262.imageshack.us/img262/2441/stereocopia.jpg

One more with the Mitutoyo, this time at 11x plus a small crop so... something like 12x, also last shot with aperture closed

Full size
http://farm6.static.flickr.com/5213/545 ... e83d_o.jpg
The stereo
http://img9.imageshack.us/img9/9762/stereo2000px.jpg

And a 40x with Nikon's M plan 40/0.50 ELWD; this time I used the 5D mkII SRaw1 mode, with an efective f40 10 mpx are more than enough. 283 shots with 0,002 steps (aproximate) ISO 400

Full size
http://farm6.static.flickr.com/5291/545 ... 2aa3_o.jpg
The stereo
http://img402.imageshack.us/img402/7672 ... eo2000.jpg
And the stereo 100% crop, I love the way the polem texture shows
http://img232.imageshack.us/img232/595/ ... eoreco.jpg
Regards
Javier

rjlittlefield · Post by **rjlittlefield** » Mon Feb 14, 2011 4:35 pm

Interesting textures and colors!

The stereo looks pretty flat to me. It clearly has some depth, but not as much as I would expect. What separation did you use?

--Rik

seta666 · Post by **seta666** » Mon Feb 14, 2011 4:42 pm

rjlittlefield wrote:Interesting textures and colors!

The stereo looks pretty flat to me. It clearly has some depth, but not as much as I would expect. What separation did you use?

--Rik

At first I thought the same but I can clearly see the depth on the ocelos, polem cluster and specially on the lower eye. That bug area is quite flat anyway. I used +2 -2 for the stereo
When is it best to use low numbers like +2 -2 and when to use big separation like +5 -5?
I may do a new stereo and see what happens
Regards
Javier

rjlittlefield · Post by **rjlittlefield** » Mon Feb 14, 2011 10:10 pm

seta666 wrote:When is it best to use low numbers like +2 -2 and when to use big separation like +5 -5?

For this stack, I guess +5 and -5 would be good.

There is no perfect rule for choosing shifts. Bigger shifts give more depth, so generally speaking it is good to use the biggest shifts that still give clean images. The biggest shift depends on several things, especially number of frames in the stack. Deeper stacks can tolerate bigger shifts.

One good rule is to look at this number:
PixelsShiftPerFrame = ImageWidth * (Percent/100) / (NFrames-1).

The limit on shifting is reached when this number is around 1/4 to 1/2 pixels shift per frame, at whatever size the images are displayed. If the number gets much bigger than that, you may see smearing.

In your case, the stereo images are displayed at 800 pixels wide. With your 2 percent shift, I compute PixelsShiftPerFrame = 800*(2/100)/(165-1) = 0.1, which is well under the limit. Using 5 percent shift instead, we would get PixelsShiftPerFrame = 0.24, which is still safe.

Going bigger than plus and minus 5 percent shift (10 percent difference) causes too much difference between the two images, and the pair gets hard to keep fused.

I think that picking best shifts is still more art than science. I will be interested to see what a new stereo looks like.

--Rik

seta666 · Post by **seta666** » Tue Feb 15, 2011 3:08 am

Thanks Rik, I am working on a new stereo. There are things with th formula not clear to me. When you say pixel with you mean a single image or the whole stereo. Also, I did the stereo of small size for web purposes; If I use full size it is more than 4000 px wide (for a single image) so; 4000(2*100)/(165-1)=0.48
Do I have different stereos for web size and for print size?
This is the new version with + - 5% shift; I find them pretty similar (Now that I replaced the wrong stereo it clearly shows more depth)
http://img254.imageshack.us/img254/3250 ... 2000px.jpg (I replaced the deffective stereo with the rigth one)
Regards
Javier

rjlittlefield · Post by **rjlittlefield** » Tue Feb 15, 2011 11:07 am

Javier,

The new version at http://img831.imageshack.us/img831/8649 ... 000px5.jpg is not stereo -- it is two copies of the same view, side by side. The two sides are not identical, however. There has been some manual editing done in two patches, one at center left of frame, the other in lower left. Perhaps you selected the wrong two images when assembling the pair?

Do I have different stereos for web size and for print size?

Usually not. Setting a smaller shift will reduce the apparent depth, which is generally a bad thing. Smaller shift may also improve the appearance of fine details, but probably one should not be studying a synthetic stereo pair for fine details anyway.

PixelsShiftPerFrame = ImageWidth * (Percent/100) / (NFrames-1)
There are things with the formula not clear to me.

Me too!

In the formula I gave, the word "pixel" really means something like "circle of confusion". If the shift between adjacent frames in the stack is too large compared to the circle of confusion, then the final image may look smeared.

When images look sharp at the level of individual pixels, then the circle of confusion is about 1 pixel and the formula makes sense as written. But for prints, the circle of confusion is typically around 1/1000 of the image width regardless of how many pixels are involved. In that case, it would be better to use ImageWidth=1000, so the result ends up quite similar to your larger web version at 800 pixels per view.

Maybe it would not be crazy to just assume ImageWidth=1000, and do some algebra to get a rough estimate that Percent ~= NFrames/20

This would say that for stacks containing more than 100 frames, you can get away with shifts up to 5%, but for 50 frames, only 2.5%.

In practice, I am not sure that any formula helps a lot, because of differences between stacks. For me, values below +-1% usually do not give good stereo separation, and values above +-5% have too much difference to view easily. So the right value lies in a fairly narrow range, and maybe the formula is too rough to help very much.

I generally find the best separation by experiment. I make a run of say 11 frames from -5 to +5, with Preferences > Preprocessing > Image Pre-sizing set to 50%, 33%, or even 25% to make the processing go faster. Then I poke through the set using Tools > Stereo > Start Preview until I find a pair that has good separation and does not look smeared. If I already have enough resolution I'm done; otherwise I make a final run at full resolution using the separation determined by the experiment. Sometimes the best pair turns out to be say -3 and +2, in which case I will "split the difference" and make the final run at -2.5 and +2.5.

--Rik

seta666 · Post by **seta666** » Tue Feb 15, 2011 1:47 pm

rjlittlefield wrote:Javier,

The new version at http://img831.imageshack.us/img831/8649 ... 000px5.jpg is not stereo -- it is two copies of the same view, side by side. The two sides are not identical, however. There has been some manual editing done in two patches, one at center left of frame, the other in lower left. Perhaps you selected the wrong two images when assembling the pair?

This mean that apart from going crazy I am also losing my sigth; I thougth first version looked better. Belive it or not I could see depth anyway (twisting your eyes does funny things to your brain)

I did the wrong stack; I entered +5% and +5% instead of -5%

I am redoing it right now

Before I used to do always +3 -3 but If the subject is deep enough I consider +2-2 to be enough; may be flat subjects need more separation

Regards and thank you for all the advice
Javier

rjlittlefield · Post by **rjlittlefield** » Tue Feb 15, 2011 2:45 pm

seta666 wrote:Before I used to do always +3 -3 but If the subject is deep enough I consider +2-2 to be enough; may be flat subjects need more separation

This all sounds right. One good source of examples is HERE -- the examples I posted when introducing synthetic stereo. But I see those examples range all over the place, from -1 +1 to -5 +5. There is no obvious relationship between magnification, stack depth, and the percent that I ended up using because it looked good.

--Rik

seta666 · Post by **seta666** » Tue Feb 15, 2011 3:39 pm

Very interesting post that one; I replaced the wrong stereo with the good one. It clearly has more depth now but to me the old one looks more "natural"
Also, here is a 100% crop were you can clearly see the texture in the polem
http://img814.imageshack.us/img814/339/ ... ecorte.jpg
Regards

rjlittlefield · Post by **rjlittlefield** » Tue Feb 15, 2011 5:29 pm

seta666 wrote:It clearly has more depth now but to me the old one looks more "natural"

This is one of the advantages of knowing the actual subject. I prefer to see more depth, but if the subject really is flat, then more depth will not be natural.

Also, here is a 100% crop were you can clearly see the texture in the polem
http://img814.imageshack.us/img814/339/ ... ecorte.jpg

The crop works very well. 3D arrangement of pollen around the ocellus is clear, and the reflection of the objective appears below the surface of the ocellus as it should be.

--Rik