基于小波系数相关性的纹理图像快速修复算法

Fast Texture Image Completion Algorithm Based on Dependencies Between Wavelet Coefficients

为了实现纹理图像的快速修复,在传统样本块修复算法的基础上,提出了一种基于小波系数相关性的图像修复算法.简要论述了纹理图像各小波变换系数的相关性特点;根据图像分解后小波系数高频成分较少、主要信息均集中在低频分量以及各分量对应位置的信息具有一致性的特点,算法首先对破损图像进行小波分解,然后利用最优样本块匹配的方法对低频分量进行修复,同时实现其他分量对应位置信息的修复,最终通过小波合成完成纹理图像的修复.仿真实验结果表明,在修复效果无明显差异的基础上,该算法比原方法的处理时间大为缩短.

According to the traditional examplar-based method, a fast texture image completion algorithm based on dependencies between wavelet coefficients was presented in this paper to complete texture images quickly. A simple analysis was made of dependencies between wavelet coefficients of texture images. Since most information about texture images focuses on low-frequency component and all components have dependencies in their corresponding place, the damaged image was first decomposed into four components by wavelet transformation, the low-frequency component was then completed by selecting the best matching patch, the other three components with their own information in the corresponding place were simultaneously completed, and the texture image was finally completed by composing all the completed components together. Simulation results indicate that the running time of the proposed algorithm is remarkably reduced, while the texture completion effect is similar to that of the traditional one.