Research On Texture Synthesis And Transfer

Texture synthesis and transfer are research hotspot of computer graphics and computer vision. They are widely used in virtual reality, computer vision and so on. Texture synthesis technology which proposed to solve seams has more applications. As seam aliasing and discontinuity exist in randomly puzzle, and image processing technology is developed quickly, research work on texture synthesis has made great progress. This paper works on previous studies of texture synthesis and transfer, make further study.Minimum error cut algorithm is the core of block-base texture synthesis. We calculate cumulative errors of overlapping area between two blocks, and pick a best line to separate them. The cumulative error on this line is minimal and it's the best result. We analysis the process of duplicated calculating which results to slow down synthesis speed, and an optimized algorithm is proposed. To the block which is patched, all of the cumulative errors of blocks are restored to minimum error map, and then the duplicated calculating is avoided. During texture patching, restricting acceptable error range will get different results of texture synthesis. In optimized algorithm, the relative error between front and back blocks are limited to get steady and smooth results.First, this paper reviews the development history of texture synthesis and transfer. Then a classic algorithm of block-based texture synthesis and transfer is introduced. This paper analyses the core, future and shortcut of original algorithm. Finally, an improved method is proposed.The main work of this paper can be summarized as follows:1. Based on block-based texture synthesis algorithm, an improved texture synthesis algorithm is proposed.2. A mapping data structure of minimum error is introduced to overlapping area in minimum error cut algorithm. The process of texture matching in texture synthesis is improved.3. Parameters selection of acceptable error range in texture synthesis is modified to optimize the selection process and improve quality of result.4. Based on block-based texture transfer, an improved texture transfer algorithm is proposed.5. After research on parameters for iterative process, each parameter is optimized.6. A non-iterative texture transfer algorithm is proposed to speed up texture transfer process.7. Black-points percentage line and black-area number line of both original texture and results after binarization are proposed as a base to evaluate synthesis results. Experiment results show the feasible of optimized algorithm in improving synthesis speed, result and evaluating quality of synthesis result.