Seeding Texture Synthesis Algorithm

In order to accelerate the speed of texture synthesis and improve the quality of results, the seeding texture synthesis algorithm is proposed in this paper. First, we pre-compute samples to obtain seeds of samples and basic attributes, then seed by using stabilization neighborhood relations and the candidate set in the synthesized space. After the completion of seeding, all seeds begin to grow. Finally, we amend the result by using of neighborhood matching strategy. Empirical results show that the seeding texture synthesis algorithm can produce the results more similar to the sample. The algorithm is valid and efficient, can complete in milliseconds and be easy to promote the parallel implementation.