建立了一类新的图像修复算法,即自适应混合图像修复算法。这一算法的基本原理是:首先将图像分解为几何轮廓和纹理两部分分别修复后再合成。对几何轮廓部分,用自适应整体变差方法进行修复,自动根据区域特点选参修复,在保持边缘信息同时加快了修复速度;对纹理部分,用二阶共生矩阵合成方法进行修复,使修复效果更为理想。
An adaptive mixed image inpainting algorithm is presented. With this method, an image to be inpainted is decomposed into two components: one is the geometric outline component and the other is the texture component. The geometric outline is inpainted by a so-called Adaptive Total Variation on method, choosing different parameters according to different region, not only preserving sharp edges and also accelerating the inpainting process. The texture component is inpainted by a so-called Co-occurrence Matrix synthesis method, making the result more accurate.