The Research And Application Based On Texture Sythesis From Sample Image

Texture synthsis from sample is a new texture synthesis technology which is developed in recent years. It's used to solve the problems of joint,distortion, and parameter adjustment which appear in the conventional method. It shows a widespread application prospect in image editing, damage image inpainting, data compression, network data transmission, large-scale production as well as rendering of reality graphics and so on. It is one of the reserh hot spots in computer graphic, computer vision and image procession.In this paper, we have researched three classic texture synthesis algorithms, analysis their respective characteristics and inadequates. And design an improved algorithm based on the combination with their advantages. In addiction, have researched the application of texture synthesis in digital image inpainting, according to the inadequacy of the existing algorithms, proposed an inproved algorithm. The main research contents in this paper bellows:(1) Research the application of texture features analysis based on MRF model in texture synthesis. Mainly reseached the WL,Ashikhmin and Efros01 ect three claasic texture synthesis algorithms based on MRF model, describe the three algorithms's implementation steps in details,analyse their principle, discuss their respective characters.(2) Based on the existing texture synthesis algorithms, design and implement an improved algorithm according to relation of texture field and matching point, proved the effction of the improved algorithms by experiment.(3) Discuss the three main parameters (block size, overlap size, sample number)'s effection to texture synthesis quality and speed.And by analyse the experiment results, get the parameters setting suggestion according to different demands(4) Research the TSFS's application principle in image inpainting, based on the existing algorithm, describe the algorithms's details implementation steps, discuss inpainting order's effection to inpainting result. (5) Guide the search of matching blocks by texture segment pretreatment, and apply the new inpainting order to improve the existing algorithm. The experiment shows that, the improved algorithm is better than the existing algorithm when remove the object or fill the damage region.