Noise image restoration based on mathematical morphology

Traditional restoration approaches for blocky noise images have the disadvantages of making the image be fuzzier, or losing the useful information. According to the particularity of blocky noise, a new restoration approach is proposed based on mathematical morphology including dilation, erosion, open and close operations. Firstly, the suspected noise is determined after the color classification. Secondly, the suspected noise region is eroded and filled by other region. Finally, a dilation operation is applied to the image. Some noise radar images are utilized to test this approach: the experimentation indicates a better restoration result in comparison with the traditional approaches. In addition, discussion is introduced to the running complexity of this algorithm.