A texton-based texture synthesis algorithm is presented in this paper. Based on the Markov Random Field (MRF) model, the algorithm takes a block randomly from a given sample texture to form a seed area in the center of the synthesized output texture, then synthesizes the output image pixel by pixel spirally along the edge of the seed area. When sampling a pixel from the sample texture in the synthesis procedure, the algorithm adopts so-called candidate search, global search and central search to preserve the integralities of textons in the synthesized output texture. It was practically proved that the algorithm does better than the existing methods in quality for synthesizing most kinds of textures, especially for synthesizing the complex textures.
展开▼
