Effect of Window's Shape on Median Filtering

Abstract

Impulse noise filtering plays important role in image processing applications and median filtering is widely applied to perform such task. The median filter's ability to preserve edges and simplicity with implementation has made it popular for the suppression of impulse noise. However, the challenge that most users of the median filtering have to deal with is the nature of the shape and size of a window to be used for the filtering. Various works on effect of window sizes have been done, but that of the shapes have not been extensively explored. This paper therefore assessed median filter's performance based on window shapes and sizes. Nine (9) different shapes of windows, five (5) different window sizes (3×3, 5×5, 7×7, 9×9, 11×11) and five (5) different noise density injection were used to evaluate the performance of the median filter. The experimental results show that the amount of noise in an image and selected window size strongly affect the output of a particular shape and that, a window's shape may not necessary produce excellent results at all times. However, the Square, Cross, Horizontal, Z-Shape and re-Z-Shape exhibited stronger performance as compared with the rest of the shapes.