Abstract :
A new image filter based on rough sets theory and multi-directions difference image is proposed for impulsive noise to overcome the drawback of median filter and its extends. The conceptions, the number of directional signs and the number of directional extremum signs, are defined to describe the smoothness and noisiness of the pixels according to the multi-directions difference image. An image is partitioned into the smooth-noise sub-image, the impulse-noise sub-image and the unpolluted sub-image according to the smoothness and noisiness of pixels by rough sets theory. The corresponding filters are selected for the two noise sub-images. The experiment results, compared with the other filters, shows that the new filter is the best for the performance, MSE, PSNR and ISNR.
