为了去除红外图像中的噪声，提出了一种基于全变分理论的去噪算法。该方法继承了经典全变分模型在去除噪声中保护边缘的优点，结合图像平滑扩散原理，得到了一个全新的扩散函数；同时引入了一个边缘检测算子，对正则项和忠诚项的相关参量进行了改进，使得修复后的图像大大避免了阶梯效应；最后对该算法的实现进行了推导。结果表明，该算法能够有效地去除噪声，并且避免了阶梯效应的产生。

In order to remove the noise of infrared images, an image denoising method based on total variation was proposed, which inherits the advantage of edge preserving of the total variation(TV)model. A novel diffusion function was derived from the theory of smooth diffusion. Simultaneously, an edge detector operator was introduced to improve the related parameters of regularization term and fidelity term, which makes the denoised image avoid the staircase effect to a great extent. Finally, the implement of the algorithm was derived and the experimental results demonstrated the superior performance of the modified denoising algorithm.