Modified Geometric Mean as an Estimator of Outlier based Artifacts in Natural Images

The Modified geometric mean is proposed as a detector for the removal of high-density salt pepper noise and impulse noise and artifacts like in images. The proposed algorithm initially finds a 3*3 pixel size. In this proposed filter modified geometric mean will act as a measure of central tendency to estimate whether the processed pixel is corrupted or not. If the pixel processed is corrupted then the median of the neighborhood is also checked for noise. If found true then the corrupted pixel is replaced with a modified geometric mean else median of the confined neighborhood pixel is unaltered if found not corrupted. From the results, we found that the proposed filter would eliminate fixed valued outliers, random valued outliers and impulsive artifacts that occur in images due to ageing. The proposed work shows extremely good Quantitative and Qualitative results. In comparison with other algorithms blurring of an image is less and it has high PSNR and SSIM and low MSE. The proposed work shows exquisite noise elimination functionality with more suitable edge preservation functionality.