A Non Adaptive Scheme for Removal of Image Noises and its Artifact

In this paper, a fixed 3×3 window size switched algorithm for removal of the image noises such as fixed impulse, random impulse noise, zero mean Gaussian noise, mixed noise and artifacts such as blotches, stripes, streaks for gray scale images for increasing noise densities is presented. The proposed method uses a tree based method to classify the noisy pixels and noisy free pixels. Being a switched algorithm the median is calculated only after identifying the pixel is noisy. At higher noisy environments the corrupted pixel is replaced either by the median or by mean of non noisy pixel in the current window or the preprocessed neighborhood pixel based on the classification. The quantitative and qualitative performance of the proposed algorithm outclasses many of the standard filters in terms of different artifact removal with edge preservation. The proposed algorithm found to give better results in the removal of fixed impulse noise up to 80%, and preserves edges up to 70%, zero mean 0.9% variance Gaussian noise. The proposed algorithm even found to suppress mixed noise of 10% impulse noise plus zero mean 0.3% variance Gaussian noise composition.