基于块拼接的纹理图像合成原理, 研究了螺旋状搜索路径搜索匹配纹理块以提高纹理图像合成速度。此外, 还通过不断增加样本空间的纹理匹配块数量, 来提高匹配的质量。通过在纹理块的采样空间、搜索方式、相似度方面展开研究, 提出了一种新的纹理图像合成算法。根据输出图像产生的新纹理, 把不断合成的输出图像也作为样本空间的一部分, 通过不断增加样本空间纹理匹配块数量, 来提高纹理块匹配的质量。在样本空间搜索纹理匹配块时, 采用螺旋状的方式进行搜索, 大大减少了搜索时间, 使合成速度得到了很大提高。在判断两个匹配块相似度方面, 同时把纹理块的颜色和梯度作为两个匹配块之间是否相似的重要参数。经过实验验证, 与传统算法相比, 本文算法在提高合成速度的同时也提高了合成质量。

Based on the principle of block splice texture image synthesis, spiral searching path to search matching texture blocks for improving the texture image synthesis speed is studied. In addition, the matching quality is improved, by increasing the number of sample space texture matching blocks. A new texture synthesis algorithm is proposed through a study in sampling space, search methods and similarity of texture block. According to new texture generated by output image, synthesized output images are seen as part of the sample space to enhance the texture blocks probability of matching, by increasing the number of texture matching block in the sample space. The spiral searching manner is used to reduce the search time greatly and improve the synthetic speed. The gradient and color of texture block are taken as the similarity parameters between the two matching blocks, for identifying the similarity degree of two matching blocks. Compared with the traditional algorithms, the proposed algorithm can improve the synthetic speed and synthetic quality with the experimental verification.