Fusion coding method of image local feature semantic distribution structure and sample distribution structure

Abstract

The invention discloses an image local feature semantic distribution structure and sample distribution structure fusion coding method. Firstly, the multi-scale dense SIFT local feature of the image is obtained, and then the semantic center of the local feature is obtained by K-means clustering, and the χ 2 The measurement measures the semantic distribution structure of local features; the Hausdorff distance is used to measure the distribution structure of image samples based on the local features of the image; the distribution structure measurement of image samples is based on the bag-of-words model encoding. Finally, through the method of matrix spectrum optimization and solution fusion structure to obtain local low-dimensional compact local features, use K-means clustering to obtain new semantic centers of local features, and perform feature quantization and coding on each image to obtain strong classification and discrimination ability image feature description. This method not only considers the distribution structure between semantic centers of local features, but also considers the distribution structure between image samples, and unifies the above two structures into one framework, so as to obtain image structure information fusion with compact local features and strong classification and discrimination capabilities. coding.

Description

technical field [0001] The invention belongs to the technical field of video surveillance image processing, and in particular relates to a fusion coding method of a semantic distribution structure and a sample distribution structure of local features of an image. Background technique [0002] In recent years, there have been more and more applications of intelligent monitoring systems based on content analysis. In order to intelligently analyze and identify targets, image description and cognition are important issues to be solved, because image local descriptions show structural diversity and each structure The complexity of describing the fusion makes it difficult to fully describe the essential structure of the image distribution considering a single structural constraint. In this case, the current method cannot carry out the fusion encoding of the semantic distribution structure of the image local features and the sample distribution structure, and thus cannot represent ...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide an image local feature semantic distribution structure and sample distribution structure fusion coding method, which solves the problem that the existing coding method cannot accurately represent the image content information

Images