Distortion level analysis of a 2D median filter with a weighted central element

Abstract

The paper presents the evaluation of the median filters with weighted central elements use for cleaning images from impulse noise. The influence of obtaining undistorted pixels probability and distortion of the image pixel distribution function on the processing is shown. It is demonstrated that median filters with a window 3 × 3 allow obtaining a higher quality of image cleaning from noise with a small probability of impulse noise, and median filters with window sizes 5 × 5 and 7 × 7 allow to obtain a higher quality of image cleaning from noise with a high probability of impulse noise. The obtained results can be used in medical image processing and adaptive filtering systems for photo and video data processing.