Abstract: In order to denoise impulse noises in images, an adaptive window switching median filtering method is proposed in this paper. First, each pixel is classified as signal or noisy point by the BPnet. Next, the image is filtered by an improved median filter. According to the result of noise detection, the filter can adjust adaptively windows width and sample choicely, each noisy point in image is denoised by filtering. In terms of suppressing impulse noise and preserving image details, the proposed method is better than that based on median filter. Even in the extreme case of 70% noise corruption, noisy images can be effectively recovered.

Key words: image denoising, impulse noise detection, adaptive switching median filters