Studies On Example-based Texture Synthesis Method Of Two-dimensional

Example-based texture synthesis algorithm by the given small texture samples can synthetictexture of any size. This method ensures the synthesis of texture images is continuous andsimilar in the visual to some extent.This thesis makes some new studies on example-based texture synthesis algorithms basedon the predecessors and presents some new understanding over the original algorithms,mainly the following points:1. On the basis of the original algorithms, this thesis introduces the research background andsignificance of texture synthesis, discusses their current research status, introducesthe existing problems and trends of present algorithms, studies some classic point-basedsynthesis and block splicing algorithm,summarizes the advantages and disadvantages of eachalgorithm and texture type which they are fit for dealing with.2. By studying and summarizing the advantages and disadvantages of Wei&Levoy algorithm,this paper proposes new improved algorithm based on L neighborhood size selection of Wei&Levoy algorithm. Through analyzing detail feature’s proportion similarity and histogram’ssimilarity of the texture image and its sub-scales, this paper presents an L neighborhood sizeadaptive selection algorithm. On block matching in texture, due to localized reasons of MFR,the regions having been synthesized can impact on regions which will be synthesized.Through spliced analysis of synthesized block and surrounding blocks which will besynthesized, to each block generating from texture samples, this algorithm first finds its set ofadjacent blocks mosaic texture, then according to the records of these can match compatiblepieces and then these records can be compatible with the block matching, selects a suitablesynthetic texture blocks according to synthesis sequence Synthetic order when calculating,finally gradually extends the generated target texture. This algorithm improves synthesisefficiency and synthesis quality of texture synthesis algorithm based on Markov random fieldmodel.The simulation and results analysis demonstrate the feasibility and effectiveness ofthis algorithm.3. This paper studies drawing the application of image restoration based on texture synthesis,and proposes a new improved algorithm using the texel size of sample used to determine thesize of matching blocks to repair the damaged areas based on Criminisi algorithm. This algorithm makes the matching more accurate according to the texel size of the sample todetermine the size of matching blocks and makes up the structure fracture defects when usingCriminisi algorithm to repair by multi-level synthesis to improve restoration resultsmulti-level effects. Experiments show that the algorithm has good inpainting results.4.Through the thoughts on the poisson equations and block texture synthesis,and base on thestudy of the characteristics of the image restoration algorithm, the original structure breaksdown into sub-sub-map and texture map two parts, then repair it according to eachcharacteristic.The simulation and results analysis demonstrate the feasibility of the twoalgorithms.For structural sub-image, repaired using methods based on the Poisson equation,for sub-graph of the texture, repaired using texture block synthesis. Experiments show that thealgorithm has good inpainting results.