Based On The Color Information Scalarizing Edge Extraction Method

Edge is one of the important features of the image, which is the most significant part oflocal brightness changes in image. Through the research on edge, we can deal with differenttypes of images or there different requirements. Color images contain more information thangray image, and the edge extraction of color image is an important content in the study of colorimage processing, but the color is a vector, there has been a concern that how to turn it into areasonable scalar. Therefore, the work of this paper has carried on the beneficial attempt. Thispaper first introduces people’s research on early gray-scale image edge detection algorithm,analyzes and compares the classical detection operator, and presents the detection results of lenaimage with appropriate parameters; then simply lists some commonly used color space, andintroduces part of the existing color image edge detection method, such as expansions toclassical operators based on a certain color space, wavelet multi-scale detection of true colorimages, algorithms combined with neural network are also researched. The paper’s concreteworks are as follows：1、It proposes a new color image edge detection method using color triangle, in whichvector is replaced by scalar using area. In this algorithm, change of color informations in thepixels are replaced by change of area of the triangles, and we transform the color informationswhich can satisfy the condition into the sides of the triangle, use one edge detection operator todefine area, then determine the edge information through threshold comparison.It considerate thelink between the color components comprehensively, and also turn the interactions of abstractvector into the calculation of the scalar, which has a strong theoretical significance, and the testresults can tell that its edge localization is accurate and clear.2、Considering cellular neural network, it uses the triangle color information as input, todetermine whether the network is convergent through convolution, after determine theimage boundary, and then locate the edge. Experimental results show that the edge informationobtained by this method is abundant and accurate, at the same time, it eliminats the noise.3、After analyzing the information quantity of the triangle, it proposes a method based onthe information quantity of the right angled trapezoid, which makes up the defect that part of theedge is missing caused by some of the color component of the pixels can’t constitute a triangle.The test results show that, compared with the algorithm based on the information quantity of thetriangle, it eliminates some confusing edge, the results are more clear.