中文摘要：

自适应总体变差算法可以解决总体变差算法无法实现去噪的同时保留图像的细节纹理。对该算法的扩散系数研究表明当算法满足如下性质：Ф’（｜grad（I）｜）≥0，Ф＂（｜grad（I）｜）≥0；当扩散系数法向量方向的值为Ф’（｜grad（I）｜）／｜grad（I）｜，扩散系数切向量方向的值为Ф＂（｜grad（I）｜），并且｜grad（I）｜→＋∞时，向量方向的值趋干0，法向量与切向量的比值趋干0；当｜grad（I）｜→0，法向量方向和切向量方向的值均大于0时可以取得较好的试验结果，并认为最小曲面函数是合适的扩散系数。

英文摘要：

Though the adaptive total variation presented algorithm a solution to preserving the detailed texture in image when denoising which the total variation algorithm failed to do, the diffusion coefficient has not been further studied. The calculation in this paper shows that when certain properties are satisfied the algorithm can be implemented with a better experimental outcome in denoising. It was concluded the minimal surfaces function be the appropriate diffusion coefficient.