采用CDD模型的自适应图像修复算法

Adaptive image inpainting algorithm based on CDD model

在CDD模型基础上,提出了一种自适应图像修复算法,引入在不同曲率选择不同修复模型的自适应系数q,使得在大曲率时使用CDD模型,其他时候使用TV模型,大大减少了CDD模型的修复时间;引入在不同变化程度选择不同扩散方式的自适应系数p,使得在破损区图像边缘较多,即大梯度时使用接近TV的模型,而在平坦区,即小梯度时使用接近热扩散的方程,使得修复效果更佳。实验证明该算法在修复速度和修复效果上都要优于CDD模型。

An adaptive image inpainting algorithm based on CDD model is proposed.An adaptive coefficient q is introduced to choose the right inpainting model for different curvatures.The coefficient adopts CDD model only for large curvatures,and use TV model for others.Thus the repair time of CDD model is greatly reduced.Another adaptive coefficient p is introduced to choose the right diffusion way for different degrees of change.This coefficient adopts a more closing TV model in damage area edge which has large gradient,and use thermal diffusion equation in flat areas which has small gradient.Then better inpainting results can be achieved.Experiments show that the proposed algorithm has much faster inpainting speed and better inpainting results than CDD model.