Discrete Minimum Distortion Correspondence Problems for Non-rigid Shape Matching

摘要: Similarity and correspondence are two fundamental archetype problems in shape analysis, encountered numerous application computer vision pattern recognition. Many methods for similarity boil down to the minimum-distortion problem, which shapes endowed with certain structure, one attempts find matching smallest structure distortion between them. Defining structures invariant some class of transformations results an or similarity. In this paper, we model using local global structures, formulate problem as binary graph labeling, show how different choice invariance under various classes deformations.