FREQUENCY DOMAIN SHAPE DESCRIPTION METHOD USED FOR IMAGE MATCHING, RECOGNITION AND RETRIEVAL

摘要

A frequency domain shape feature description method used for image matching, recognition and retrieval relates to technical fields of pattern recognition, image processing and computer vision, comprising: dividing an image area into several sub-areas, using each feature point as a reference point respectively to collect statistics on a number of other feature points in each sub-area so as to obtain a histogram, calculating and using a power spectrum of the histogram as a shape description, and calculating a similarity between shape descriptors of two images to obtain an image similarity, thus implementing the image matching, recognition and retrieval. According to experiments, a shape context spectrum feature is provided with better rotation and telescopic invariance, robustness against noise and distortion, and a smaller calculation overhead.