Methods And Applications Of Fast Texture Synthesis Based On MRF Processing

Texture synthesis is the hotspot research in current computer graphics and computer vision. Texture Synthesis From Sample(TSFS) is one of the new techniques for texture generation emerged in recent years, dividing the sample into small pieces, which can be described as a two-dimensional Markov model. TSFS can overcome some defects of Texture Mapping, such as texture seams and texture distortions. Moreover, it can avoid the boring processes of parameter optimization of Procedural Texture Synthesis. In recent years, TSFS has become one of the research hotspots in computer graphics and computer vision.In the first chapter, the related concept of texture synthesis techniques and the development of research background and the status quo at home and abroad are briefly introduced. For MRF algorithm we approach two-pronged research from a single sample map and diverse plans, summarizing the existing texture synthesis algorithm based on the shortcomings. Then we presented the content, methods and ideas in this paper.In the second chapter, MRF algorithm is divided into pixel-based, block-based collage and mixed-synthesis. We discussed about the existing classical algorithm in detail and then summarized for the advantages and disadvantages of each algorithm. Meanwhile we discussed about varieties of texture types, and subdivided the structural texture and the applicability of the classic algorithms.In the third chapter, from the free parameters on the starting blocks of texture synthesis, we analyzed and discussed the synthesis of synthetic texture quality and speed of change, proposed two kinds of improved algorithms, which is L partition and the variable neighborhood block under the control of texture synthesis method, reducing the free parameters on the synthesis of quality and speed of impact. In the first improved algorithm, for the two-factor of neighborhood width and block size infect periodic structure texture Synthetic, we use split L neighborhood, to speed up the process of pretreatment to reduce the traditional L-neighborhood in the whole process of matching search existence of computational complexity and matching errors.And using the second matching method to improve the speed of synthesis. For fixed-size block matching errors, we use texture synthesis of variable block under the control of using self-adaptive to adjust the size of the texture block. According to the threshold or not to change the texture block matching size, thus to obtain the optimal matching block texture block, thus to eliminate the need for cumbersome manual adjustment of the texture block. The use of block suture patterns reduces the non-continuity of the block joints.In the four chapter, we presented multiregional constraint-based diversity texture synthesis method for the diverse map texture synthesis' poor results by algorithm's complexity and choice of algorithms. The algorithm changes the constrained texture synthesis algorithm multiple map generation process, separate the synthesis process and constrain process. So it overcomes the variety of map texture synthesis repeated to determine the tedious constraints of regional boundaries, accelerates the synthesis of speed, and increases the diversity of sample map selected,thus meets the multi-texture design in variability requirements.In the five chapter, we discussed the texture synthesis algorithms and presented the corresponding synthesis results and application examples in detail.In the last chapter, we summed up the full text, and pointed out innovation points in the paper, and proposed further research direction.