传统的总变分修复算法，在文本与划痕的去除上需要较多的迭代次数，耗时较长。针对此问题，提出一种改进算法，在正则项中加入扩散调节系数，在扩散调节系数中引入随着迭代次数而变化的自适应参数，使算法在迭代初期以较快的速度扩散，而在迭代后期减速扩散，从而使修复后的图像保持锐利的边缘。大量仿真证明，改进后的算法在划痕与文本的去除上减少了迭代次数和运行时间，修复后的图像视觉效果很好。

The classical total variation image inpainting algorithm costs more iterations and time to remove the text and scratches. Aiming at this problem, we propose an improved algorithm. Firstly, the diffusion accommodation coefficient is added into the regularization term. Secondly, a self-adaptive parameter that varies with the number of iterations is added to the diffusion accommodation coefficient. Thus, the algorithm diffuses at a faster speed in the early iteration and slows down in the later iteration, and the restored image maintains a sharp edge. A large number of simulations show that the improved algorithm reduces the number of iterations and running time in the scratch and text removal, and the restored image has good visual effect.