High Density Fixed Valued Impulse Noise Removal Using Improved Decision Based Hybrid Median Filter and its Application on Medical Images

Abstract

Formation of image is affected by image capturing device characteristics and intensity of light. Therefore inferior quality of image capturing device and inadequate lighting conceals particulars and significant details associated with image. In order to squeeze out the hidden features, image enhancement (noise removal) is mandatory. Hence noise removal is realized as pre-processing step in image study. In this research paper a Decision Based Hybrid Median Filter is suggested for the renovation of gray scale images corrupted by fixed valued Impulse noise. This suggested filtering technique offers superior results than the earlier known enhancement techniques like standard median filter (MF), adaptive median filter (AMF), fast and efficient median filter (FEMF), new adaptive weighted mean filter (AWMF), noise adaptive fuzzy switching median filter (NAFSMF). The foremost objective of the suggested technique is to enhance visual perception and boost peak signal to noise ratio (PSNR). In the suggested technique if current pixel is found noisy then it is substituted by median value achieved after eliminating pixels with intensity values 0 and 255 from the window. When all the pixels in the window are 0 and 255 then the window size is increased. If 80 to 90% pixels are 0 and 255 in the new window, then substitute the pixel with the previously processed pixel value else substitute it with mean of the designated window. The suggested technique is verified against standard as well as medical images. Comparisons and experimental results shows better visual and quantitative results with reduced mean square error (MSE), increased image enhancement factor (IEF) and peak signal to noise ratio (PSNR).