The Application Of Spectral Graph Wavelet Transform In Image Denoising

With the development of science and technology,the image information occupies the mainstream for people to obtain information.People have higher and higher requirements for image quality.However,in the process of image trapsmission and storage,the image information will be disturbed by a variety of noise.The noise interference causes image distortion.In order to obtain more information from a noisy image and meet the need for more advanced image processing,the image must be-denoised and the denoising effect will directly affect the subsequent image processing.The concept of our proposed algorithm is to take advantage of denoising performance of the BM3D algorithm and the edge preservation property of spectral graph wavelet transform based image processing.The main work of this paper includes the following aspects:1.The research background and significance of image denoising,the current situation of image denoising at home and abroad,and the quality evaluation methods after image denoising were introduced in detail.Spectral graph wavelet transforms and patch-based denoising algorithms were in-depth analysis of research.2.Research and analysis of spectral graph wavelet transform and the design of two-channel wavelet filter banks on graphs,For many image denoising algorithms,there is a common edge blurring problem.Spectral graph wavelet transform algorithm can effectively detect images in irregular domains.The characteristics of the edge information are applied to the image denoising algorithm to better protect the image details.3.The principle and method of BM3D algorithm analyzed emphatically,and a new image denoising algorithm combining the spectral graph wavelet transform and BM3D algorithm was proposed.The similar patch was obtained through the patch matching algorithm and constructed a graph by connecting obtained patches,then,which the spectral graph wavelet filter bank based on the graph signal was applied to remove Gaussian noise for better preserving the edge information of the image.Experimental results indicated that,compared with wavelet transform denoising algorithm,non-local mean denoising algorithm,and BM3D denoising algorithm,the denoised images obtained by the improved algorithm presented in this paper show better denoising results in both subjective evaluation and objective evaluation and have significant performance improvement in image edge detail preserving.