Fusion Edge Detection Method Based On Canny Operator And Morphology

In the image analysis and processing, the role of the edges is important. But,the difficulty of edge detection is filtering noise while enhancing edges. This paperresearches the traditional Canny operator and morphological algorithm respectively,and a new edge detection algorithm is obtained, specific work as follows:Firstly, by using the traditional Canny operator, edge detection exists twoproblems. One is that, the filter can not eliminate the interference of local noise, andin the detection of pseudo-edges while losing edge detail. The other one is that, thefilter is hard to set Gaussian filter parameters for different images lackingadaptability. In this paper, the effect of image edge detection by noise issignificantly reduced through an improved Canny edge detection operator, by usingadaptive median filter instead of a Gaussian filter. Meanwhile, it maintains the edgestrength and has more complete details, gets a more satisfactory subjective effects,as well as significantly improvement in objective evaluation.Secondly, if the concentration of noise at a high leve, the effect of detection isnot ideal, by using traditional morphological algorithm for image edge detection.Therefore, this paper proposes an improved algorithm of morphology to resist thenoise, which uses two different structure elements to edge detection. For the imagewith the salt and pepper noise improved algorithm can filter the noise effectively, theedge location of the improved image is accurate, more smooth and clear outline, theinformation contained is also more abundant. Meanwhile, the subjective andobjective evaluation results have improved.Finally, for the improved Canny operator, peak signal to noise ratio andstandard deviation are much better. Moreover, the effect of the correlationcoefficient of improved noise immunity morphological, the degree of distortion andthe average gradient are much better. Therefore, integrating the two improvedmethods, the obtained edges image shows that the information entropy, root mean square error, correlation coefficient and the degree of distortion are better than thetwo improved methods, visual effect is also satisfactory.