An improved directional difference filter for impulse noise reduction

Directional difference filter is a nonlinear filter to reduce the impulse noise. It does not employ order statistics such as median filter but employs directional difference in the local area. As it considers the edge information in the local area, it can reduce the impulse noise with preserving the image information compared to the other methods. However, it still requires a noise detector as pre-processing to avoid the image blur. In this paper, we look to the concept of a nonlinear filter labeled epsiv-filter and propose an improved directional difference filter, which can reduce the impulse noise with preserving the original image information without pre-processing. We also show the experimental results to evaluate the performance of the proposed method compared to other methods.