Displacement estimation, stereo matching and "object" recognition - a computer simulation approach working with real world imagery

This paper presents a method for displacement estimation in images, for stereo image matching and the recognition of image "objects" that has been developed for industrial real time applications but with biological notions in mind. The first stages of the method are compatible to biology (lateral interactions, threshold- ing), well understood in terms of signal processing, robust in the presence of noise or inadequate signal conditions and work with almost every type or real world imagery. This method might therefore serve as a test hypothesis of how early vision could work.