使用中值-各向异性扩散的超声图像去噪算法

摘要

针对超声图像的散斑噪声,提出一种基于多方向中值滤波和改进各向异性扩散的去噪算法.该算法利用多方向中值滤波的良好边缘保持能力,在滤除噪声的同时注重边缘细节的保持.使用归一化局部方差和图像梯度组成的扩散系数,避免了传统各向异性扩散算法中梯度阈值为常数带来的鲁棒性差等问题.通过多组仿真实验,综合滤除散斑噪声能力、保持边缘能力以及迭代速度等指标,表明该算法比传统PM和SRAD模型有更好的滤除超声图像噪声和边缘保持能力.%A denoising algorithm using multidirectional median filtering and improved anisotropic diffusion was proposed for ultrasound images corrupted with speckles. Multidireetional median filtering was used to remove the speckles and preserve edge details by its promising edge-preservation capability. A diffusion coefficient incorporated with unitary local variance and image gradient was introduced to obviate poor robustness introduced by the constant gradient threshold in the traditional anisotropic diffusion algorithms. Compared with Perona-Malik model and Speckle Reducing Anisotropie Diffusion model, experiments show that the proposed algorithm has better performance for noise removal, edge-preservation, and iteration speed.