The Algorithm Of Image Denoising Based On SVM And Non-local Means

Now, digital image is the base of a lot of information in the life, however, it will be influcnced by noise when it is transmitting. So the quality of image will be down and perhaps lose the detail of image, it becomes very important for us to denoise.This paper focuses on support vector machine image denoising method and nonlocal means denoising technology. The main work of this paper includes:1. A Nonlocal-means algorithm for image denoising based on wavelet momentIn order to denoise well, we improved Nonlocal-means properly and took wavelet moment instead of traditional method of using pixels to compute similarity. So we proposed a Nonlocal-means algorithm for image denoising based on wavelet moment.Firstly, the image was decomposed in different subbands of frequency using the nonsubsampled wavelet transform. Secondly partitioning the coefficients with the fixed size and then using the wavelet moment to compute the similarity of pixels. Finally the modified nonsubsampled wavelet coefficients were transformed back into the original domain to get the denoised image.2. The image denoising algorithm using nonsubsampled contourlet transform based on least squares support vector machineIn order to improve classification, using the least square linear system for function based on support vector machine, so we proposed the image denoising algorithm using nonsubsampled contourlet transform based on least squares support vector machine. Firstly the image was decomposed by nonsubsampled contourlet transform in order to get coefficients. Secondly constructing the preliminary binary map by thresholds and using the coefficients to construct feature vector with the space rules, and then putting them into least squares support vector machine for training. Lastly the thresholds were computed in different frequencies and orientations to denoise. The modified nonsubsampled contourlet coefficients were transformed back into original domain to get the denoised image.3. The image denoising algorithm using curvelet transform based on fuzzy support vector machineDue to classification performance of support vector machine, importing the fuzzy function based on support vector machine, so we proposed the image denoising algorithm using curvelet transform based on fuzzy support vector machine. Firstly the image was decomposed by curvelet transform in order to get coefficients. Secondly constructing the preliminary binary map by thresholds and using the coefficients to construct feature vector with the space rules, after that putting them into fuzzy support vector machine for training. Lastly the modified curvelet coefficients were transformed back into original domain to get the denoised image.