螺旋线轨迹在纹理合成算法中的应用

本文在MRF模型和已有块拼贴算法的基础上,根据纹理的邻域相关性以及匹配点的位置分布在上次合成匹配点邻域的概率较大这一特点,直接选取前一匹配块的相邻块为待选最佳匹配块,使算法能够快速地采样到最优匹配块,并能较好的保持合成纹理的邻域相关性;同时,采用螺旋顺序逐块合成纹理,能够加强对合成过程的约束,克服了合成过程中不连续区域的出现。本文算法较好的维持了样本图的结构特征,在保证纹理合成质量的同时,能够有效的减少算法的计算量。

Application of Spiral Path in Texture Synthesis Algorithm

In this paper we present an algorithm based on MRF model and some previous patch- based algorithms. According to theprobability that matching pixel locality distributes near to the last best- matched pixel is higher and the neighbor correlation of textureimage, we choose the neighbor patch of the last best- matched patch in the sample image as the preparative best- matched patch di-rectly. So this algorithm can find the best- matched patch faster and keep neighbor correlation of output texture better; On the otherhand, we synthesize textures patch by patch spirally. This method can enhance the restriction for the synthesizing procedure, andavoid the discontinuous area happened in the synthesized image. It was practically proved that the modified algorithm can keep betterstructural characteristic of sample image in synthesized image, and produce similar results while the computation is reduced.