3d macro

[Clintonwake]

Hi All

Is there a way or what technique/process/software is there to convert image to 3D ("rocking")?

I searched here and on the net and do see some example where 3d specs are not required but the image just rocks from side to side a few cm.

cheers
Denis

[rjlittlefield]

I'm not sure exactly what you're asking about, but I'll try to answer anyway.

Stereo and rocking are two different ways of presenting almost exactly the same information.

In stereo, the subject is shown from two different angles at the same time, routing only one angle to each eye using some sort of viewing device or special viewing technique such as crossed eyes. The viewer's brain extracts 3D information by matching features in the two images and figuring out from the "disparity" which points are closer and which are farther away.

In rocking, the subject is also shown from multiple angles, but one at a time in temporal succession. The viewer's brain extracts 3D information by tracking the positions of features over time and figuring out from the relative motions which points are closer and which are farther away.

From a production standpoint, one big question is how the two or more views are collected.

In the world at large, most commonly the subject is photographed from each angle separately.

However, under some circumstances it is possible to shoot a single "stack" (as for focus stacking) and then render it from a variety of angles. The resulting views can then be used to produce synthetic stereo and/or rocking. See HERE for some discussion. Roughly speaking, the process is limited to stacks that contain at least 20 frames and angles that lie within the cone of light that was originally accepted by the camera lens. This means it works much better at high magnification, where the cone of light is wide and the DOF per frame is disproportionately shallow, leading to deep stacks. This is the same situation that makes it awkward to physically shoot two stacks from different angles, so this is one of those rare processes that works best exactly when you need it the most.

Most of the stereo and rocking that you see here at photomacrography.net was produced using the latter process, working from a single stack.

Once the multiple views are available as single images, then the other big issue is a matter of formatting to present them as either a static stereo pair, or an animated GIF or looped movie for rocking, or even combining the two techniques to produce a rocking stereo. Numerous different software tools can be used for that process.

Is the answer to your question in there somewhere?

--Rik

[Clintonwake]

Thanks Rik,

My apologies for the lack of information as I did not know how to actually convey my question.

Yes, you have answered my question perfectly and I did purchase Zerene a year or 2 ago but was not aware of that feature even though I went to the Zerene website. I did notice that Helicon, searching the net last night had that feature.

I encourage all my friends learning macro to join this group, as apart from the wealth of information here, the members are so so helpful, credit to you all.

thanks once again
cheers
Denis

[rjlittlefield]

was not aware of that feature even though I went to the Zerene website.

Hhmm... Thanks for that feedback. Synthetic stereo and rocking is mentioned on the basic How To Use It page, but maybe putting another list of features in the Tutorials section would help.

I did notice that Helicon, searching the net last night had that feature.

Helicon's feature is superficially similar but actually quite different when you go to use it. What Helicon does is to infer a continuous 3D "terrain surface" (no overlaps, no cliffs). Onto that they map the RGB pixel data, and then they re-render the result from whatever angle you want. Their approach has the advantage that for sufficiently simple surface shapes they can move a lot farther off-axis. But "sufficiently simple" means things like the stamped face of a coin. If you try it with something like an insect's body, or worse, anything with bristles, then you get results that I publicly describe as "less than completely realistic". Zerene's method can work straight from the images, without attempting to infer 3D structure. This restricts the angle, but gives much cleaner handling with complicated geometry like overlaps and bristly bugs. You can see an example at http://www.photomacrography.net/forum/v ... hp?t=13302, 4th panel. There's a very recent example of Zerene's stereo at http://www.photomacrography.net/forum/v ... hp?t=24192. For illustration I tried running that same data through Helicon Focus 6 just now, but it seems to have some sort of resource problem and refuses to show the RGB data -- just gives me a white surface when I ask to do that. I can see from the depth map that it fails to capture the structure of a sparse array of hairs sticking out from the eye, but unfortunately I can't show that fact here.

--Rik

[Clintonwake]

Thanks Rik

Having it in the tutorials would be good.

Thanks for the examples and coincidentally I purchased the Camranger for my stackshot. Going to try that this weekend.

cheers
Denis