Binary segmenting tree based regional feature description method and system

Abstract

The invention is applicable to image feature extraction, and provides a binary segmenting tree based regional feature description method. The regional feature description method comprises the steps of A, cutting an original image into multiple segmented regions; B, calculating similarities of adjacent segmented regions, and combining two adjacent regions with the highest similarity into a super-region; C, repeatedly performing the step B until a complete original image is combined; D, establishing a binary segmenting tree according to the segmented regions and the super-region; and E, extracting features of a target region and the features of the super-region corresponding to the target region from the binary segmenting tree, and establishing region features based on the extracted features, and performing target regional feature description according to the regional features. According to the regional feature description method, the basic regional features of the target region, the equative super-region, the adjacent superior region and the super-region are combined for describing features of the target object; and therefore, local region features are included while overall situational information is included as well, so that the binary segmenting tree based regional feature description method has the advantages of situational perceptivity, high recognition and high robustness.

Description

technical field [0001] The invention belongs to the fields of computer vision and image processing, and in particular relates to a method and system for describing region features based on a binary segmentation tree. Background technique [0002] Feature description is a basic module in computer vision and image processing. In applications such as image salient object detection, object segmentation, object recognition, and semantic image retrieval, the robustness and richness of features largely determine its performance. Most of the feature extraction methods currently used are: the original image is divided into image sub-blocks or image regions by image segmentation algorithm or pixel clustering method, and then image features are extracted according to these image sub-blocks or image regions. [0003] Although the existing feature extraction methods are efficient, natural images often contain a lot of complex backgrounds, and the most advanced image segmentation methods ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a region feature description method and system based on a binary segmentation tree, which aims to solve the problems of poor feature robustness and insufficient information when performing feature extraction in the prior art

Images