Project 3, Stereo

For this project, you should implement the Cooperative Stereo algorithm 
of Zitnick and Kanade.   It is clearly described in the publication:

A cooperative algorithm for stereo matching and occlusion detection
C. Zitnick and T. Kanade
IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, 
No. 7, July, 2000, pp. 675 - 684

Which can be downloaded from here.

Try running your algorithm on the popular stereo images from 
http://middlebury.edu/stereo/  
Notice that these are rectified images, thus the motion of a point 
between the two images will be completely horizontal.

For output, please compute depth images for the 4 sequences from the 
Middlebury site.  (Tsukuba, Sawtooth, Venus, Map)

(By a "depth image" we just mean an image where the intensity of each 
pixel is proportional to the disparity of the corresponding point.  Don't 
worry about using the same scaling factors as in the results section of 
the middlebury site.  This is the standard way to view output from 
stereo algorithms.)

Discuss how well you found your algorithm to work.  In particular, you 
will notice that the algorithm propagates match information over depth 
discontinuities.  Can you think of any way in which to improve this?

Web Accessibility