摘要： 基于马尔可夫随机场(MRF)模型和已有的块采样合成技术，提出一种改进的合成算法，采用图论的最小割线算法对MRF的似然性进行优化，将合成时的光栅扫描方式改为之字形扫描方式，以抑制纹理的倾向性和纹元的不完整性。试验证明，该算法的合成效果较好。

关键词: 纹理合成, 马尔可夫随机场, 最小割线, 最大流

Abstract: This paper proposes a new patch-based sample texture synthesis algorithm based on Markov Random Field(MRF) model and other related algorithms. In the new algorithm, the min-cut algorithm of graph theory is adopted to optimize the likelihood of MRF and improves the scanning method from raster scanning to zigzag scanning. The new scanning method can restrain texel’s imperfection and texture’s orientation.

Key words: texture synthesis, Markov Random Field(MRF), min-cut, max-flow

中图分类号: 

• TP391