Based On The Kind Of 2d Texture Synthesis Method

Virtual reality is a complicated and multiple technique based on CAD/CG, multimed-ia and artificial intelligence thus remains an important research field. There are great improvements in image based rendering (IBR) technique recent years. More details that can not be acquired by geometry models now can be rendered by IBR methods. Texture synthesis is very important in realistic rendering based on images and has become one of the focuses of virtual reality, computer graphics and image processing.Texture synthesis can be classified into two types. One is procedural texture synthesis which directly stimulates the natural forming procedure of objects on the surfaces of the objects, such as hair or wood texture. Those kinds of methods create textures vividly, but for each kind of texture, all the parameters must be adjusted by interaction and then textures being re-rendered repeatedly. Furthermore this kind of algorithms can only describe some kinds of texture not suitable for prevalent textures.The other kind of methods for texture synthesis is based on the model of Markov Random Field. This kind of methods is a new direction that has been improved very quickly these years. The MRF model supposes that textures have strong local features based on statistics. That is any pixel of texture is decided by its surrounding neighborhood This kind of methods for texture synthesis pieces up texture patches which are perceptually continuous on surfaces based on a small sample texture image and the object geometry model.Based on the MRF model my thesis research focused on the texture synthesis methods in 2D and put forward the following new thinking and algorithms:1. Many of the recent texture synthesis methods select an output pixel by searching with its already generated neighborhood for a corresponding...