Improving performance of K-SVD based image denoising using curvelet transform

Image denoising algorithm in transform domain which uses learning of dictionary has better PSNR performance than others. It is seen that the popular algorithms based on K-SVD proposed earlier has still in use. However, the texture part of the image could not be preserved during the process of denoising. It is also seen that the effect becomes more visible with increased value of standard deviation of the Gaussian noise. The proposed algorithm in this work uses curvelet transform along with K-SVD to retain the texture part of the image. The denoising with the proposed method shows better PSNR performance as compared to denoising with only K-SVD.