Speckle Reducing Anisotropic Diffusion Based on Directions of Gradient

In this paper, we focus on the problem of speckle removal and edge preservation by means of anisotropic diffusion in ultrasound images. The conventional anisotropic diffusion is performed in four directions (north, south, east, west) without condition in the previous methods (Perona and Malik, 1990; Black et al., 1998 ; and Hae Yong Kim, 2006). In this paper, a new anisotropic diffusion method based on directions of gradient is proposed. At first, is calculated by using GHAD (gradient hostogram-based anisotropic diffusion) for anisotropic diffusion. Secondly, Prewitt mask is performed for the magnitude of gradient and directions of gradient. As a result, when the magnitude of gradient is large enough, the pixel is considered as the edge. In the case of edge, the weight of diffusion is selected adaptively according to the directions of gradient. Otherwise, the proposed diffusion is performed at 8-directions (north, south, east, west, north_left, north_right, south_left, south_right) which are not an edge. Experiments results show that the proposed method can eliminate noise/speckle and can remain contour of edge in ultrasound images with speckle noises.