Study On Natural Texture Image Synthesis

Texture-image synthesis is the important task of Digital Image Processing and Computer graph. It can be abroad used into communication and security and others, so it has high value.At present, it is not mature. So now it doesn't satisfy the demand of multifarious texture synthesis. The primary task of this question for discussion is to make use of the more mature theory of artificial intelligence into texture synthesis. At last, it is carried out successfully by using ANN arithmetic. The ANN BP arithmetic can solve the problem of nonlinear mathematics effectively, and possess higher collateral processing ability.The task mostly includes three sides: at first, we study the eigenfunctions of various textures to describe these character from every sides and select 16 eigenfunctions; secondly, we set up a ANN BP model by way of the numerical value of 16 eigenfunctions as its inputs, then don't circulate the model until it is steady; finally, it will form the software of texture synthesis. Giving a texture image, it can create a similar texture automatically. The experiment shows that this method has the excellence such as short-time, clear and so on. In addition, it can apply various texture images.