Abstract: The selectivity of diffusion direction and the ability of directional diffusion are the main characteristics of anisotropic diffusion filter. Efficient expression of local feature of signal or image is the base of anisotropic diffusion. Traditional gradient expression is always disturbed by noise. This paper defines an intension parameter of image local anisotropy and proposes a novel diffusion model by analyzing image local structure expression manner. The results of tests indict the proposed method possesses better abilities of de-noising and directional diffusion, compared with ALM model and CAZ model, and signal-noise ratios improves by 0.1 dB~0.8 dB and 0.3 dB~1.2 dB respectively.

Key words: anisotropic diffusion, local structure of image, image de-noising and enhancement

摘要： 各向异性扩散平滑去噪的主要特点是扩散方向的选择性与定向扩散能力，有效表征信号或图像的局部结构特征是各向异性扩散的基础，传统的梯度表示方法极易受到噪声干扰。该文在分析图像局部结构表征方式的基础上，定义一个图像的局部各向异性强度参数M，提出一个新的扩散方程。试验测试显示，新扩散模型相对于ALM模型与CAZ模型具有更好的噪声压制能力和定向扩散能力，信噪比分别提高了0.1 dB~0.8 dB和0.3 dB~1.2 dB。

关键词: 各向异性扩散, 图像局部结构, 图像去噪与增强

CLC Number: 

• TN911.73