A Measuring and Adaptive Detection Method of Image Edge Gray Scale Fluctuation

Abstract

The invention provides an image edge grey level fluctuation measurement and adaptive detection method so as to overcome the defects that the threshold value of an edge detected by a Canny edge detection method needs to be manually set, and theoretical bases are in shortage. Firstly, new image grey level fluctuation measurement-normalization gradient intensity is defined, and it can be proved that normalization gradient intensity of an edge position and normalization gradient intensity of a uniform image region are subject to noncentral chi2 distribution and chi2 distribution with the freedom degree being 2. Secondly, based on statistical distribution characteristics of the normalization gradient intensity, a high edge detection threshold value and a low noise false alarm rejection threshold value are adaptively set through a hypothesis testing theory, and then edge constant probability detection and noise constant false alarm rejection are achieved. Finally, a magnetic lag threshold value for edge detection is formed by combining the two threshold values, and then adaptive detection can be achieved for the image edge. Edge constant probability detection and noise constant false alarm rejection can be achieved, and experiments show that the detected edge is excellent in performance and high in robustness and automation degree.

Description

technical field [0001] The invention relates to the technical field of image analysis, in particular to an image edge gray scale fluctuation measurement and self-adaptive detection method. Background technique [0002] Edge is the most basic and effective feature to describe the contour shape of an object. In an optical image, it usually corresponds to the position where the geometry, physical properties or external environmental factors of the object change, such as the depth of the scene, the surface normal vector, the material, the temperature and the lighting conditions. , which has objective physical meaning and contains the shape information of the object of interest. Edge detection can not only extract the original shape features of the object, but also greatly reduce the amount of data that needs to be processed for subsequent image analysis. [0003] Due to the extreme importance of edges in object shape description, edge detection has always been a research hotspo...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the deficiency that the threshold value of the Canny method for edge detection needs to be manually set and lacks theoretical basis, the present invention proposes a measure of image edge gray level fluctuation and adaptive detection in order to improve the automation degree of the Canny edge detection method and improve the edge detection performance. method

Images