Abstract: A solid texture synthesis algorithm is proposed to synthesize layered textures which have varying statistical features along the depth direction. It employs the global optimization-based synthesis framework. In preprocessing, a feature distribution map that reflects the varying statistical features along the depth direction of the sample is generated, and it is involved in the similarity measurement. During synthesis, the neighborhood similarity in three orthogonal directions is considered, and k-coherence search is utilized in the iterative optimization. Experimental results show that the algorithm can produce high quality results which bare reasonable feature distributions both along the depth direction and in horizontal slices, and it converges quickly.

Key words: layered texture, solid texture synthesis, feature distribution map, k-coherence search, iterative optimization, sample

摘要： 针对深度方向上具有特征差异的层式纹理，提出一种基于样图的体纹理合成算法。采用基于全局优化的合成框架，在预处理中生成反映样图深度方向上特征差异的特征分布图，将其纳入相似性度量，合成时考虑3个正交方向上的邻域相似性，并利用k相关性搜索进行迭代优化。实验结果表明，该算法能保证合成纹理在深度方向上特征分布的相似性及水平截面的特征合理性，具有较高的合成质量和较快的收敛速度。

关键词: 层式纹理, 体纹理合成, 特征分布图, k相关性搜索, 迭代优化, 样图

CLC Number: 

• TP911.73