Research on High Precision Algorithm of Scale-Spatial in Computer Vision

Abstract

Abstract The intra class differences of object categories will bring great challenges to the image matching algorithm in machine vision. In order to guarantee the accuracy and quality of image matching algorithm, it is necessary to accurately define and understand the scene and semantics of the image, so as to find out the typical similar features of the image. On account of this, this paper first analyses the role and traits of scale-spatial theory, then studies the high-precision algorithm of scale-spatial in computer vision, and finally gives the extremum detection of scale-spatial, the determination of extremum position and the generation of feature point descriptor.