Application Of Graph Wavelet Transform In Image Segmentation

Image segmentation occupies a pivotal position in the field of digital image processing.Image segmentation is to segment and extract the target part of interest, which is the basis for subsequent image recognition and analysis. There is a broad application prospects of image segmentation. Up to now, a variety of imagc segmentation methods have been developed, but there is still no one generally applicable to all images. The traditional wavelet transform has better local characteristics only in the regular domain-time domain and frequency domain. In this paper, we extend the scope of the study to the irregular domain-graph domain, and use the Graph Wavelet Transform(GWT) method to segment the image.This paper mainly completes the following work:(1) According to reading a lot of literature, understanding the progress of image segmentation at home and abroad, and analyzing the advantages and disadvantages of different image segmentation methods, image segmentation is carried out by using the Graph Wavelet Transform and Haar xwavelet.(2) Through simulation experiments by MATLAB, the same image were segmented by Sobel, LOG, Canny and GWT four algorithms, and the experimental results were compared and analyzed.(3) The two-dimensional discrete wavelet transform(2D-DWT) is used to segment image to the same image, and the experimental result is compared with that obtained by using the GWT algorithm.(4) By setting different ? values and noise levels respectively, the same image is segmented by using the GWT algorithm, and the experimental results are compared and analyzed.Through simulation experiments by MATLAB, the experimental results show that compared with Sobel, LOG, Canny and 2D-DWT, the GWT algorithm can not only detect the edge information of the image effectively, but also has better robustness to the noise.