基于各向异性散布的医学图像非线性滤波法

A Medical Image Nonlinear Filtering Method Based on Anisotropic Diffusion

为了向临床医生提供清晰准确的诊断依据 ,在对医学断层图像 (CT、MRI)进行滤波处理时 ,须保留具有重要诊断意义的微细结构。然而 ,绝大多数滤波技术在去噪的同时却滤出了细小结构。本文介绍了一个改进的非线性各向异性散布滤波算法 ,图像滤波被认为是一种散布迭代过程 ,通过自动确定最优散布常数和迭代次数 ,散布过程在遇到边界时就会被抑制或停下来 ,从而保留了边缘信息和微小结构。通过对实际医学图像(CT、MRI)的实验表明 。

When medical images are filtered, fine details important to specific diagnostic tasks must be retained in order to provide doctors clean and reliable evidence. Most of conventional filtering methods enhance signal to noise ratio(SNR), while at the same time object boundaries and fine structural details are blurred. In this paper, an improved nonlinear anisotropic diffusion filtering algorithm is described. Filtering is considered as an iterative diffusion process. Because the diffusion process is suppressed or stopped at boundaries by means of automatically defining of optimal diffusion constants and iterative times, edges and small structures information are retained. Results of experiments with real medical images (CT, MRI) demonstrate that our method can improve SNR, and at the same time it can retain important anatomical details.