A New Speckle Reducing Anisotropic Diffusion for Ultrasonic Speckle
用于超声斑点噪声滤波的各向异性扩散新模型

Although speckle reducing anisotropic diffusion (SRAD) has been greatly explored in ultrasonic imaging, it generates block effects in homogeneous regions and blurs edges due to the intrinsic diffusion. To address these defects, we propose a new speckle reducing anisotropic diffuse (NSRAD) in which a new Sigmoid function as the diffusion coefficient is suggested. Due to the specific properties of such novel diffusion coefficient at various region, NSRAD is able to diffuse: 1) in a nearly constant speed within homogeneous regions, thus to avoid the block effect; 2) in a sensitively varying speed within the transition regions, at the same time, in a speed approaching to zero more quickly than that in SRAD concerning the structure regions, thus helps to improve the resolution and enhance the weak edge while sharpening the strong edges at the edges. The performances of the proposed model are verified in experiments with synthetic and clinical images. Especially, a quantitative comparison between NSRAD and SRAD is made upon simulation data, where we demonstrate that NSRAD suppresses speckle more efficiently in homogenous regions hereby has less distortion, while in inhomogeneous regions it enhances details and sharpens the edges more effectively than SRAD.