A fuzzy filter for random impulse noise removal from video

Abstract

We present a new filter for image sequences corrupted with random impulse noise. The main goal is to optimally combine noise removal with the preservation of the image details. The filtering strategy is to remove the noise in three different successive filtering steps and a fourth refinement step. In each filtering step, only the pixels that are detected as being noisy are filtered. The noise detection is achieved by fuzzy rules. To exploit the temporal information in image sequences as much as possible, detected pixels are filtered in a motion compensated way. The experimental results show clearly that the proposed method outperforms other state-of-the-art filters both numerically (in terms of the peak-signal-to-noise ratio) and visually.