Satellite Imagery
 |
The camera approach is similar to the way we make pictures of our planet. One camera (in a satellite) takes small pictures of the Earth, and those many pictures are stitched together to make a single, unwrapped image of the globe. I started by building a small turntable (Legos…is there nothing they can’t do?). |
Placing the skull on the turntable, I photographed it eight times, turning the the turntable about 30 degrees for each photo.
A quick bit of math will inform the clever that this was not a full orbit. I did this one as a proof of concept, and stopped short of the full 360.
I then brought them into Photoshop and made a single file with all eight picture stacked in layers. The base layer was the photo of the skull facing forward.
We work with two layers at a time. With the opacity of the upper layer set to 50%, move the layer around until you find the feature that most overlap, and align them. The next step is to take a large, soft-edge eraser and remove half the image. It may seem a little confusing, but remember that you are building out the mosaic image in a direction.
In this case, I moved the upper layer to the left to align the nasal cavity, so I’m going to erase the overlap on the right [Link]:
Adding several more slices [Link]:
You can see that there is some discontinuity at the top of the skull, giving the top of the skull a bumpy look, where the curvatures don’t match. These errors stem from not having enough samples.
The error is most pronounced on the top, as the top of the skull is nearly spherical, and we are essentially unwrapping it in a cylindrical projection.
The bottom of the skull is more cylindrical (at least on the front half), therefore mapping better into the cylindrical projection. If the number of photos were increased, taken in smaller increments, the distortions up top would be less noticeable.
Here’s the final unwrapped image mapped back onto a simple 3D skull shape [Link].
One problem for future experimentation is lighting. This version has some very distinct lighting present in the image map. A simple front illumination rig should help eliminate this.
Update: A better solution would be a simple PVC Light Tent. I’ve built one, and will try it out real soon now.
New Approaches
I’ve stumbled across an interesting idea that may make the head scanner idea a little easier. I need to get my hands on a second-hand hand scanner, or small photo scanner.
These are mechanically vastly more simple than their flatbed brethren, and the drivers are designed to expect arbitrary scan sizes. It’s on the to-do list, after the 3D Scanner (which I already have the parts for) is completed.
* The final result of which was a cupful of liquid that was almost, but not quite, entirely unlike tea.
File Under: How-To, DIY, Mad Science, Make, Hardware, Rotisserie Scanner
