Image Denoising Using Principal Neighborhood Dictionary Non Local Means for Color Images

This Paper presents an appealing loom to Non Local Means (NLM) image denoising algorithm that uses Principal Component Analysis (PCA) for dimensionality reduction. This Principal Component Analysis is applied individually for the three bands namely red, green, blue of the color images crooked with a frequently occurring noise model namely the Gaussian noise. As a result of PCA, obtain the Eigen value measures on the observed data and then develop a small number of artificial data set (principal components) using Parallel Analysis that will account for most of the variance in the observed data set. The principal components can then be used as predictor or criterion variables in subsequent analysis. Consequently measure the neighborhood similarity weight for the subspace using the normalization technique. The ensuing algorithm is referred to as the Principal Neighborhood Dictionary (PND) Nonlocal Means. The performance and the computational time of the proposed method is improved by the retrieval of the significant principal components from the projected subspace for each bands and increase of subspace window size respectively. The accuracy of the proposed PND method is compared with the other state of art methods and the superior performance of the proposed image denoising method is stated in terms of the increased Peak Signal to Noise Ratio (PSNR).