Abstract：The inherent speckle noise in ultrasound images severely affects the quality of images. In this paper,an algorithm based on tensor voting to reduce the speckle noise is proposed. Tensor voting presents a unified computation framework that can implement the smoothness constraint to generate structured information, such as isolated points,curves,and surface from a 2-D or 3-D image. Using this feature,the isolated points are exacted by tensor voting and the curves and surface feature are emphasized on to restrain the speckle noise. The experimental result shows that compared with other spatial filter,the algorithm based on tensor voting can more effectively remove the speckle noise and improve the quality of ultrasound images.

Keyword：tensor voting;ultrasound image;image de-noising;