The Novel Denoising Technique Based on 8D Optimal Directions Using Statistical SD for Random-Valued Impulse Noise

This paper proposes a novel denoising technique for suppressing a random-valued impulse noise. At first, the proposed noise detection is based on determining the optimum direction, by estimating the Standard Deviation (SD) in different eight directions of the filtering window. Next, the similarity index is estimated by average Absolute Difference (AD) for comparing of how close the processed pixel is to its neighbors in the optimum direction. Later, the processed pixel is noiseless if the similarity index is low value otherwise the processed pixel is noisy pixel. Comprehensive results verify that the proposed technique is exceptional efficiency than other previous denoising techniques at all noise rates.