Abstract

Image denoising algorithms usually are dependent on the type of noise present in the image. There is a great need of a more generally usable, noise independent denoising algorithm. In this paper, an image denoising technique is proposed where the image is first transformed to the nonsubsampled contourlet transform (NSCT) domain, detail coefficients are extracted and feature vector for a pixel in the noisy image is formed by the spatial regularity. The support vector machine (SVM) is then used for classifying noisy pixels from the edge related ones. Finally, the denoising is done by shrink method, where an adaptive Bayesian threshold is utilized to remove noise. Experimental results show that the method gives good performance in terms of visual quality as well as the objective metrics such as peak signal to noise ratio (PSNR).