Detection and Removal of High Density Random Valued Impulse Noise

In this paper, it has been intended to detect the random valued noise and then remove it with the approximation of neighboring pixels. The detection process comprises of two parts. One for the border detection and other for the detection of the rest of the image. In detection process median is computed taking fixed 5×5 window. A pre-defined threshold value is set for the detection of corrupted and un-corrupted pixels. In removal process row wise and column wise matrix operations are separately performed on two distinct images. The output of the above two operations are merged together to get a new matrix. Then conditional mean operation is performed to replace the noisy pixels. Lastly border removal is done and the overall image is further smoothed by unconditional mean operation. Experimental result shows that the proposed filter outperform other filters in respect of performance at noise density as high as 65%.