Fast texture synthesis based on texture standard property

By analyzing the effect of sub-patch parameters on speed and quality of texture synthesis, we find that for a certain type of images which don't depend on the parameters can obtain rapid and effective synthesis. We choose L2 distance between adjacent sub-patches as a measure, propose a new algorithm to calculate the coefficient of texture standard property. According to a large number of statistical results, we give a criterion of distinction of strong standard textures, weak standard textures and general images, and apply it to texture synthesis, use large scale sub-patch and zero search to synthesize strong standard textures. The results show that even compared to the global search methods, we can obtain rapid and effective results.