Automatic Deformable Shape Segmentation for Image Database Search Applications

摘要

A method for shape based. image database indexing is described. Deformable shape templates are used to group color image regions into globally consistent configurations. A statistical shape model is used to enforce the prior probabilities on global, parametric deformations for each object class. The segmentation is determined in part by the minimum description length (MDL) principle. Once trained; the system autonomously segments deformed shapes from he background, while not merging them with adjacent objects or shadows. The formulation can be used to group image regions based on any image homogeneity predicate; e.g., texture, color, or motion. Preliminary experiments in color segmentation and shape-based retrieval are reported.