Image Denoising Method Based on v-Support Vector Regression and Noise Detection

摘要

Aimed at the correlation between noise pixels and neighboring pixels, a new method based on the v-support vector regression (v-SVR) is proposed to remove the salt & pepper noise in corrupted images. The new algorithm first takes a decision whether the pixel under test is noise or not by comparing the block uniformity of the 3×3 window with one of the entire image, secondly adjusts adaptively the size of filtering window which is used to determine the training set according to the number of noise points in the window, thirdly determines the decision function that is used to predict the gray value of the noise pixels by means of training set, finally removes the noises in terms of the decision function based on v-SVR. Experimental results clearly indicate that the proposed method has a better filtering effect than the existing methods such as standard mean filter, standard median filter, adaptive median filter by means of visual quality and quantitative measures.