中文摘要：

提出一种新的基于纹理合成的图像修复算法,针对Criminisi算法优先权函数不准确等缺点,改进优先权函数;另外在模板的选择上,利用方差表征频率的变化来选择模板的大小,不仅可以抑制噪声对模板选择的干扰,也更能体现图像的局部差异性,从而能更加准确地寻找最佳匹配块。算法经过实验仿真证明了其准确性。

英文摘要：

A new algorithm for image inpainting based on texture synthesis is proposed,which uses a new priority function to improve the deficiency of inaccuracy the priority function of Criminisi algorithm has.Besides,for template selection,the size of template is selected according to the variation of the frequency characterised by variances,it can restrain noise＇ s impacts on template selection and can also reflects better the local differences,so that it is able to search optimal matching block more accurately.The algorithm has been verified its accuracy through the simulation experiments.