Coherence-based stereo utilizes simple disparity estimators which work
directly with the image intensities. An intensity-based algorithm might have
some difficulties with images composed only of black and white pixels, like
stereograms. But thatīs not the case for coherence-based stereo. Here's
the result of a calculation with such an image pair (see also here for results with a sparse RDS):
Another difficult test for stereo algorithms is the repetitive structure
found in many in SIRDS or Magic-Eyes. These images use the same techniques
as RDS, but utilize a single image (that's were the name comes from:
SingleImageRDS), which is repeated several times.
Wonderful SIRDS are produced by
distributed by Universal Press
Syndicate to hundreds of newspapers.
Here's an example of a disparity map calculated from such a Magic-Eye:
A Magic-Eye, coding a Turkey,...
... and the calculated disparity map
If you find other Magic-Eyes on
the web, you can try to process them at the advanced
© 1994-2003 - all rights reserved.