Print PDF
Favorite
Email
Step 2Theory
Ok, I know you don't want to hear about the theory behind this, but I'm a professor, so you're going to. Beyond that, although the method is quite simple, there are details that can make a huge difference in how effective the method is.
The key is understanding what happens to the out-of-focus (OOF) portion of an image. An image is nothing more than the sum of what the lens does to each point of light in the scene. Most people, including many otherwise really smart people writing fancy image-processing algorithms, think that an OOF image of a point of light looks like a Gaussian blur -- but that's not how lenses work. For a typical well-corrected lens, the image of an OOF point light source is actually a bright disc whose sharp outline is shaped like the lens aperture. In fact, there are a couple of instructables that take advantage of this fact to make OOF points take-on interesting shapes, giving lenses very distinctive bokeh.
Technically, the shaping is caused by hard-clipping of the point spread function (PSF) of the lens by the aperture. That implies the light near the edge of the OOF point's image actually came through the lens near the corresponding edge of the aperture. Since opposing edges of the aperture have a distance between them, all we need to do is distinguish rays coming through near the leftmost and rightmost edges -- we can capture a stereo pair in a single shot with a single lens! A larger distance between the edges allows a greater interocular distance and a more extreme stereo effect.
I've been working on methods that can computationally perform this separation for well over a year... it's a very hard problem. However, it isn't hard to separate the two views by imposing a special, color-coded, aperture. An appropriately-coded aperture can directly produce the stereo pair encoded as an anaglyph.
In fact, J. D. Songer's 1973 patent, #3,712,199, teaches coloring of the halves of an aperture to capture anaglyphs. The discontinued and rare Vivitar Series 1 Qdos 70-210mm lens even implemented this trick using a special segmented internal filter. The method described in this instructable is conceptually very similar, but is much simpler and cheaper to implement -- and it captures higher-quality anaglyphs.
The key is understanding what happens to the out-of-focus (OOF) portion of an image. An image is nothing more than the sum of what the lens does to each point of light in the scene. Most people, including many otherwise really smart people writing fancy image-processing algorithms, think that an OOF image of a point of light looks like a Gaussian blur -- but that's not how lenses work. For a typical well-corrected lens, the image of an OOF point light source is actually a bright disc whose sharp outline is shaped like the lens aperture. In fact, there are a couple of instructables that take advantage of this fact to make OOF points take-on interesting shapes, giving lenses very distinctive bokeh.
Technically, the shaping is caused by hard-clipping of the point spread function (PSF) of the lens by the aperture. That implies the light near the edge of the OOF point's image actually came through the lens near the corresponding edge of the aperture. Since opposing edges of the aperture have a distance between them, all we need to do is distinguish rays coming through near the leftmost and rightmost edges -- we can capture a stereo pair in a single shot with a single lens! A larger distance between the edges allows a greater interocular distance and a more extreme stereo effect.
I've been working on methods that can computationally perform this separation for well over a year... it's a very hard problem. However, it isn't hard to separate the two views by imposing a special, color-coded, aperture. An appropriately-coded aperture can directly produce the stereo pair encoded as an anaglyph.
In fact, J. D. Songer's 1973 patent, #3,712,199, teaches coloring of the halves of an aperture to capture anaglyphs. The discontinued and rare Vivitar Series 1 Qdos 70-210mm lens even implemented this trick using a special segmented internal filter. The method described in this instructable is conceptually very similar, but is much simpler and cheaper to implement -- and it captures higher-quality anaglyphs.
| « Previous Step | Download PDFView All Steps | Next Step » |
 |
Add Comment
|
