超声图像斑点抑制和边缘增强的异质扩散方法

证明了非锐化掩模滤波等价于具有Huber流导函数的PM异质扩散.基于最小可信尺度和局部尺度控制,提出一种用于对数压缩超声图像斑点噪声抑制的异质扩散(LSCAD)方法.该方法无需假设斑点统计分布,直接从一、二阶方向导数的局部特征估计噪声强度,并利用所估计的局部噪声设计可调扩散过程,能根据图像中斑点程度在任何位置或方向上停止.合成图像和临床B超图像实验证实,本法能有效抑制斑点噪声、增强组织边缘而不丢失特征细节.同时,将该方法与前人方法进行比较以验证其性能.

An Anisotropic Diffusion Method for Speckle Reduction and Edge Enhancement in Ultrasound Images

The existed unsharp masking speckle suppression filter was approved to be a form of Perona-Malik anisotropic diffusion with Huber conductance function. A local scale controlled anisotropic diffusion (LSCAD) method for speckle reduction in Log-compressed ultrasonic images were presented. The main idea comes from minimum reliable scale and local scale control. Without assuming the statistical distribution about speckle, this method directly estimates noise strength from local feature of the first and second-order directional derivatives. Furthermore,the local noise estimation is used to design an adjustable diffusion process,which can stop at any location and orientation according to speckle extent. The synthetic and clinical images experiment shows that the LSCAD method can efficiently reduce speckle and enhance edges, while preserving resolvable details. In addition,the results from previous techniques were compared with the proposed method to demonstrate its performance.