Adaptive Edge Detection Based on Multiscale Wavelet Features

This paper presents a novel edge detection method based on multiscale wavelet features and genetic fuzzy clustering algorithm (Yunying Dong et al., 2005), which can perform image edge detection in an automatic way. Firstly, an effective feature extraction algorithm using wavelet transform is proposed to extract classification features, thus the feature vector for each pixel is gained, which contains the gradient information in various directions; and then, these vectors are used as inputs for the genetic fuzzy clustering algorithm, which result in an automatic classification; finally, make a binary map according to the classification results, and the obtained binary map is the edge map we obtain by proposed method. Some comparisons with classical edge detection algorithms are given in this paper. Experimental results demonstrate the effectiveness of the proposed method. In addition, due to the multiscale wavelet features, the proposed method has better visual quality than the other edge detection algorithms