Synthetic utilizing differential and filter operator to construct WNN for image edge detection

Abstract

This paper constructs a Wavelet Neural Network (WNN) for image edge detection based on the fact that image edge detection is essentially a classification problem and WNN has powerful classification and identification capacity. The innovations of this paper include utilizing information of differential and filter operators in constructing the network and integrating the advantages of Canny and LOG operators in the selection of network training samples. Experimental results indicate that the method proposed in this paper can extract the image edge information effectively and the network presents good generalization ability. Also, a new edge detection based on wavelet transform of modulus maxima threshold is also proposed in this paper during the research on WNN. Finally, the selection of threshold, wavelet function and other parameters is discussed of WNN.