An improved median filtering for impulsive noise cancellation [image denoising]

Summary form only given. In this paper, an improved approach to remove the impulse noise from highly-corrupted images is presented. When the image is severely corrupted, the widely-used median filtering approaches may fail to give satisfying results. For the sake of eliminating the noise on the image data with ease and efficiency, a two-stepped filtering method, comprising outlier information detection and dependable information determination, is proposed. By comparing the considered pixel on the corrupted image with that on the median-filtered image, the proposed method aims at reconstructing the original image and preserving the details as well as possible. Through experimental illustrations and comparisons, with the proposed scheme, the capability for detail preservation and the remarkable performance of high probability noise reduction are verified.