传统的各向异性扩散滤波方法都是从偏微分方程本身出发的,理论上的分析较为复杂.本文研究了基于图像特征方向的内在正交坐标系,分析了在此框架下的扩散滤波机制,然后直接从该坐标系下建立各向异性扩散滤波方案.这样的扩散滤波方法更加直观,可以简化理论分析.在此框架下,提出了一种新的各向异性扩散滤波方法.数值实验结果表明,新的扩散滤波方法可以更好地考虑图像的局部特性,从而完成细节保护和噪声消除的双重功能.所以,基于图像特征方向建立的各向异性扩散滤波方法更能达到我们预期的效果,该设计方法是有效的.

Traditional anisotropic diffusion methods always come from PDE(partial differential equation) itself.Which results in complex theoretic analysis.The inner orthogonal coordination based on the directions of the image feature is discussed in this paper.Subsequently,anisotropic diffusion methods based on the inner orthogonal coordination can be constructed directly.Which are more intuitionistic than the traditional methods,and can simplify the analysis in theory as well.A new anisotropic diffusion filtering method which is based on this framework is proposed in this paper.The numerical results show that the new method takes the local feature of the image into good account and can accomplish both detail-preserving and noise-removing.The anisotropic diffusion methods based on the direction of the image feature proposed in the paper is efficient and can get anticipate results.