Research On Color Image Edge Detection Algorithms

With the rapid development of science and technology,digital image processing has become an important research area in signal processing.The edge,as one of the most basic features of the image,is of great significance to human visual perception.It not only outlines the basic outline of the target object,but also can convey the most of the important information in the image.The edge detection technology having been the research hotspot of digital image processing,are widely used in license plate recognition,intelligent transportation,remote sensing,mapping and many other fields.In recent years,a large number of researchers have done research on color image edge detection and obtained some research results.However,it is still to be further explored in terms of how to extract weak edges efficiently and accurately or improve the anti-noise performance of existing algorithms,and so on.This thesis based on the national natural science foundation of China(61502219),research on the color image edge detection related issues.The main contents are as follows: 1.The existing color component output fusion methods can't make a better balance with the anti-noise performance and the edge detection accuracy,so a fuzzy edge detection algorithm based on robust principal component analysis is proposed.The algorithm decomposes the noisy color image into a sparse image and a low rank image by robust principal component analysis model in three channels,and then uses an improved image fuzzy edge detection algorithm to detect the edge of the low rank images.The algorithm could effectively suppress the different types of noise while improve the accuracy of edge localization.2.Aiming at the problem of partial edge information loss caused by the correlation of color components in the output fusion methods,a color image edge detection algorithm in low channel correlation color space is proposed.The algorithm constructs an adaptive color space using wavelet and principal component analysis,and then uses the improved image fuzzy edge detection algorithm in this space,to improve the integrity and accuracy of the final edge.3.In vector approaches,the performance of edge extraction in low contrast region is not effective,and the obtained edge is relatively coarse.Aiming at these problems,the quaternion is introduced into the operator of smallest univalue segment assimilating nucleus.The edge detection algorithm combining smallest univalue segment assimilating nucleus and quaternion in RGB space is proposed.The experiment results show that our method can obtain the more complete edge information,and improve the edge location accuracy without increasing the complexity of the algorithm.