基于图论优化的约束纹理合成

提出一种基于图论优化的约束纹理合成算法对图像缺陷部分进行修复.首先给出一种基于图论优化的纹理合成方法,该方法采用图论中的最大流-最小割算法对纹理块之间的匹配进行优化;然后对此算法进行扩展,在纹理合成的过程中采用不同类型边界区域来寻找匹配.使得被修复的图像区域过渡自然,没有明显边界感.

Constrained Texture Synthesis Optimized by Graph Theory

This paper proposes a technique of Constrained Texture Synthesis Optimized by Graph Theory in order to be used to repair the blemish of picture. First,a method of Texture Synthesis Optimized by Graph Theory is proposed,in which the min-cut/max-flow of graph theory is used to match the overlap among textures. Second,it extends the algorithm by means of different types of edge region to find matching in the process of texture synthesis. This method makes no obvious sense of fringe in the boundary of the repaired picture.