With some post processing (despeckling) many of the isolated bad matches can be removed to give a cleaner looking disparity map. There are still plenty of bad matches here, especially on my shirt which is black and featureless, but this sort of output could be useful for detecting people in front of the camera, doing face detection, substituting alternative backgrounds or detecting obstacles on a mobile robot. http://code.google.com/p/libv4l2cam/
Produce a Point cloud demonstration in which RGB-D (voxels) can be seen.
mojtahedzadeh77 5 months ago