Abstract
Images usually get contaminated with random-valued impulse noise (RVIN) at the time of compression, transmission and during encoding of images. The RVIN corrupts the pixels with any value in the dynamic range of pixels (for 8-bit image the dynamic range is 0 to 255). In the proposed method, a novel approach of dividing-sliding-window (DSW) into sub-windows (textons) for identification and removal of RVIN is presented. The pixels are identified as noisy pixels on the basis of four textons median values. Noisy pixels are replaced with the median value of texton having minimum absolute inner difference. Experiments have been done on the state-of-the-art images with standard RVIN methods shows that the proposed method outperforms over standard methods. DSW can preserve the fine edges and image details in better way.