Cascade window-based procedure for impulse noise removal in heavily corrupted images

摘要

A cascade of filtering windows is implemented iterativelynfor removing random-valued impulse noise in heavily corrupted im-nages. This method is based on the peer group concept (PGC), so anpixel is considered as noise-free if and only if for each window size,nthere exists a peer group of certain threshold cardinality for it. Oth-nerwise, the pixel is considered as noisy. In the restoration process,nthe corrupted pixels are restored by taking the mean value of thenremaining good pixels in the filtering window. Extensive simulationsndemonstrate that the proposed method produces competitive resultsnat low noise rates, but at high noise rates, it outperforms other state-nof-the-art methods. This approach efficiently suppresses the im-npulse noise, shows a low computational complexity, and has annequal effect on both color and gray-level images.