Texture Synthesis with Self-Correlation Distinguishing and Its Applications

Texture synthesis from sample is a most important part of computer graphics. And the key of the texture synthesis technique from samples is local texture similarity matching. In order to improve the speed of texture comparability distinguishing, a method of texture synthesis with self-relativity distinguishing is presented in this paper. With the increase in the size of neighborhood and sample, the amount of calculation in the traditional texture synthesis approach will grow quickly and the inferior position in the synthesis speed will tack on progressively. And in the neighborhood L, the distribution of pixel is consectary at the geometry but discrete at the color space. So, this algorithm sets up a simple finding-list of self-relativity distance to the sample texture. By using the self-relativity distance of pixels in the neighborhood L as the distinguishing rule, the algorithm uses finding instead of fussy calculation in the traditional approach. It reflects the inter-character and the correlation of the texture into the distinguishing rule, quickens the speed of texture synthesis, and adjusts the multiple precision texture synthesis to avoid the problem of texture joint. After the examination, the algorithm presented has vast application in texture synthesis, image repair and texture searches, and fits for the demand of real-time in application.