A Median Filter for Preserving Image Quality Based on the Analysis of Real Impulse Noises
Saem Park, Takahiko Horiuchi, Shoji Tominaga · The Journal of The Institute of Image Information and Television Engineers · 2008
Well-known denoising methods using a median filter can restore images corrupted by random-valued impulse noise. However, because such filters treat all pixels in an image the same, they often modify pixel values that are not contaminated by noise. We describe a new adaptive median filter that analyzes actual noise. Our algorithm can detect a noise component based on the noise characteristics and replace the component with an appropriate signal component. The effectiveness of our algorithm was demonstrated in experiments using actual natural images including real noise.