Hybrid image denoising using proper orthogonal decomposition in wavelet domain and total variation denoising in spatial domain

In this paper an efficient denoising technique is introduced for removal of noise from digital images by combining filtering in both the domains, the wavelet and the spatial domain. Here AWGN noise is under considered and treated as a Gaussian random variable. In this work the proper orthogonal decomposition (POD) is applied in wavelet domain which spreads the energy of the signal in to a few principal components, and noise is spread over all the transformed coefficients. Hence suitable shrinkage algorithms can be applied on the basis of proper orthogonal decomposed components also noise can be eliminated without blurring the edges. The resultant image obtained by above algorithm is denoised again in spatial domain by using total variation denoising. This hybrid processing provides better performance in terms of PSNR as compared to individual processing as shown by experimental results.