Wavelet thresholding and joint NL Means filtering

In this paper, we propose a new framework of image denoising which employs multilevel wavelet thresholding (MWT) and non local means (NLM) filtering. The given noisy image is subjected to multilevel wavelet decomposition and thresholding is applied on detail subbands coefficients in each level to remove the high frequency noise. A spatial domain NLM filtering is applied for reconstructed first level approximation subband coefficients to remove low frequency noise. Altering both the detail and approximation subband coefficients in the proposed hybrid framework gives improved denoising performance over both wavelet thresholding method and NL Means filtering. Experiment was conducted by adding Gaussian noise to standard test images and the results of denoising performance have been obtained in terms of Peak Signal to Noise Ratio, Structural Similarity Index and execution time. Experimental results show that proposed filter gives better denoising performance with respect to wavelet thresholding, NL means filtering and multi resolution bilateral filtering (MRBF) which is a similar hybrid denoising framework.