Distinctiveness Maps for Image Matching

Stereo correspondence is hard because di erent image features can look alike. We propose a measure for the ambiguity of image points that allows matching distinctive points rst and breaks down the matching task into smaller and separate subproblems. Experiments with an algorithm based on this measure demonstrate the ensuing e ciency and low likelihood of incorrect matches. 1 

• Type: Conference
• Year: 1999
• Where: ICIAP
• Authors: Roberto Manduchi, Carlo Tomasi
• Keywords: Di Erent Image, Distinctive Points, ICIAP 1999, Image Points, Image Processing