The Research Of Error Metrics And Pixel Resynthesis In Texture Synthesis

With the three-dimensional computer graphics and virtual rapid development of image technology, texture synthesis technology has become one of the hot field of computer graphics today, and is widely used in virtual reality, computer vision and other fields. Textured surface can exhibit a wealth of detail features, you can describe a wide range of natural phenomena has duplicate features. Texture synthesis technology in these fields have broad application prospects. With the rapid increase in recent years, three-dimensional computer graphics capability and strong demand for large-size high-quality texture, how to quickly and efficiently generate arbitrarily large texture image texture blocks according to a small area has clearly become one of the urgent problems.Based upon existing texture synthesis algorithm, the texture synthesis algorithms and error metrics pixel resynthesis research and analysis to improve, modify the method for selecting the optimal error metric, the target image and generates an image surface extraction and error invalid pixel detection process, while the detected invalid pixel by pixel resynthesis tree vector algorithm and construct the pixel map for traversing the late pixel composition. The main work includes:1. Details of the development process of texture synthesis techniques, and each stage of technological development is the classic algorithm in-depth exploration and elaboration.2. For the best texture synthesis error in mixing the block selection process of selecting measure the impact of fluctuations on the results of research to improve. By analyzing the input texture block image feature information, in-depth study of the various features of the selected image error metric guide, get error metrics gray value standard deviation dependencies. The final results validate different gray value standard deviation range of the image to select the corresponding error metrics has been significantly improved in the synthesis of effect and time-consuming aspect.3. Improved search-matching algorithm invalid pixel re-synthesis process. Improved scanning line search algorithm in the invalid pixel search improve search efficiency. Tree Structure Vector Quantization algorithm accelerated by constructing a binary code as a neighbor pixel value in this category for all of the same image, avoid scanning line art very poor search.4. Proposed and implemented helix pixel search algorithms to accelerate resynthesis invalid pixels in the invalid pixel re-synthesis. Introduction of helix-pel search algorithm constructing pixel traversal mapping table(Pixel Traversal-Map) in the invalid pixel around the determined invalid resynthesis strategies pixel based on the error values in the table pixels, so that the invalid pixel resynthesis efficiency is significantly improved.