Anisotropic diffusion for preserving boundary-edge

In this paper, a new approach for anisotropic diffusion based on the direction of a boundary-edge is proposed to better preserve edges along with the removal of noises. Although the anisotropic diffusion is effective in removing noises in the image, it often destructs the structural information of the object in the image because of a sequence of inaccurate diffusions. In the proposed method, prewitt mask is performed for the directions of boundary-edge. As a result, when the magnitude of boundary-edge is large enough and the number of pixels whose values are similar in the 3 × 3 pixel structure are one figure among 5,6 and 7, the center pixel is considered as the boundary-edge with accurate edge direction. In the case of boundary-edge, the weight of diffusion is selected adaptively according to the directions of boundary-edge. The proposed diffusion is performed at 8-directions(north, south, east, west, north_left, north_right, south_left, south_right) for preservation of boundary-edge and remove of noise. Experiments results show that the proposed method can eliminate noise and can remain contour of boundary-edge in images with noises.