基于样本块与曲率特征的图像修复改进算法*

针对目前基于样本块的图像修复算法在图像修复过程中容易产生错误的匹配纹理块，难以保持纹理结构连贯性的问题，提出了结合等照度线的曲率特征和高斯函数的图像修复改进算法，首先在数据项中引入了反映纹理结构特征的曲率因子来计算优先权；其次运用高斯函数更新置信项，避免了因置信项快速下降而导致的误匹配问题。通过计算修复结果的PSNR值与其他算法进行对比，实验结果表明，该算法对丰富纹理信息的图像有更好的修复效果。

Improved algorithm for image inpaingting based on sample block and curvature features

At present, the image inpainting algorithm based on sample block is easy to produce false matching texture blocks in the process of repair. Besides, it is difficult to maintain the texture coherence. In order to address this kind of problem, we proposed an improved image inpainting algorithm, which combines curvature characteristic of the isophotes and Gaussian function. First of all, we introduced curvature factor reflecting the texture feature to calculate the priority in the data term. Besides, we used Gaussian function into the confidence term updating to avoid error matching in the process of image inpainting due to the rapid decline of confidence term. Compared with other algorithms, simulation results show that the proposed algorithm has a better effect on the image which contains rich texture information according to the value of PSNR.