关键词: 小波滤波, 数学形态学, 边缘检测, 广义交叉验证

Abstract: While the traditional methods for image edge detection of images containing noises are insufficient for today’s use, this paper presents a new and better method for image edge detection which is based on wavelet filter and mathematical morphology. This method uses Generalized Cross Validation（GCV） standard to select the wavelet threshold adaptively and cleans noises contained in images by wavelet filter based on generalized threshold function of the selected threshold. The method constructs 4 representative structure elements, selects automatically the appropriate structure elements according to the direction of the edge, and uses an improved morphological edge-detecting operator to detect image edges. And it is proved that this method can give relatively perfect results of image edges’ detection in the condition of noise.The results of this experiment show that the method can effectively suppress noise and make the detected edges more consistent and clearer. Above all, the effect of the detection is much better than that of those traditional ones.

Key words: wavelet filter, mathematical morphology, edge detection, Generalized Cross Validation（GCV）