Color-Anaglyphs: Construction & De-Construction

	Viewed through a red filter placed in front of the left, and a green or blue filter in front of the right eye, this anaglyph image is perceived in 3D and in color.

An anaglyph carries the information of the left view in the red color channel, and the information from the right view in the green and blue channels of the colorimage:

An anaglyph is created by taking the data from the red color-channel of the left image, and combining this with the green and blue channels of the right stereo view.

The red filter in front of the left eye extracts only the information of the left view, and a green or blue filter placed in front of the right eye only the information from the right view. Thus, the two eyes actually see the following images:

Viewing a color-anaglyph through appropriate filters separates the stereo information for the left and right eye (the two images are arranged so that one can cross-fuse them). As can be seen, neither the left nor the right eye receives any useful color information. Interestingly enough, we are usually unaware of this fact.

If one is using good quality filters, perception of the original colors of the scene can be achieved. You can create your own anaglyphs from stereo images at this page.

The fact that we can perceive the colors in an anaglyph raises an interesting question:

Normally, it is assumed that color information is already re-coded at the retinal level into red-green and blue-yellow color channels. This would obviously mix up the stereo information. In addition, as the above image shows, the color information at the retinal level has not much to do with the colors we actually perceive when viewing a color-anaglyph. Color perception has to take place at a later stage, where the information from both stereo channels is available. But a simple combination of the color information from the two eyes would not work, due to the disparities in the anaglyph - the images donīt match.

Before color perception can be achieved, the human visual system has to fuse the data from the two eyes into the common cyclopean reference view. Only in this cyclopean view, and only after fusion, the original color-information is available again.

This remarkable operation of the human visual system can be simulated with a neural network implementing coherence-based stereo.

In such a network, the information of the left and right input channels is automatically fused into the common cyclopean view during the extraction of the 3d-data of the scene. The cyclopean view has the correct colors!

Based on this cyclopean view (now with correct color-information) and the recovered 3d-data, it is even possible to reconstruct the original stereo pair:

A stereopair reconstructed from the above color-anaglyph, and arranged here for crossed-eye fusion (which means the right stereoimage is on the left).

For comparison with the above stereo pair, which was reconstructed from the color anaglyph, here's the original stereo image from Mexico:

The original stereopair from which the color-anaglyph on top of this page was made from (also arranged for crossed-eye fusion).

If you have anaglyphs of your own, you can process these online - on this page. Check it out!

---

© 1994-2003 - all rights reserved.