Binocular stereoscopic vision image feature extraction method combining shape and color

Abstract

A binocular stereoscopic vision image feature extraction method combining shape and color belongs to the field of data processing of multimedia information retrieval, intelligent information processing, data mining and the like, and overcomes the defects of binocular stereoscopic vision image feature extraction is not high in precision and high in complexity. The method includes the steps of: performing depth map extraction on a binocular stereoscopic vision image through stereo matching; performing contour shape feature extraction on a binocular stereoscopic vision depth image based on a window of a selected size and performing dimensionality reduction; using a sliding window detection method to perform feature extraction on a complete depth image and performing dimensionality reduction; performing color feature extraction on a binocular stereoscopic vision left image to from histogram features; and performing Gaussian normalization on contour shape features and color features, thereby realizing multi-feature fusion binocular stereoscopic vision image feature extraction. The binocular stereoscopic vision image feature extraction method combining shape and color can obtain accurate and low-dimensional binocular stereoscopic vision image features, and can be well applied to content-based indexing and retrieval of related resources.

Description

technical field [0001] The invention relates to a binocular stereo vision image feature extraction method combining shape and color, which is suitable for binocular stereo vision image feature extraction, matching and retrieval, and belongs to the fields of computer image processing technology, multimedia information retrieval and the like. Background technique [0002] At present, the stereo vision industry and technology are developing rapidly. The common form of stereo vision is binocular stereo images or videos, which are associated left and right images or videos obtained through binocular stereo cameras or video cameras, usually using side-by-side Format representation. Under the network and new media environment, the analysis and acquisition of stereo vision resource content has become an important direction of demand, and it has also become one of the important fields of multimedia data application and research. The complexity of the stereo vision resource structure...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is: to overcome the shortcomings of realizing the feature extraction of binocular stereo vision images based on monocular two-dimensional images, as well as the shortcomings of low accuracy and high complexity of feature extraction, aiming at the principles and characteristics of binocular stereo vision images , introducing the depth map feature, invented a binocular stereo vision image feature extraction method combining shape and color

Images