Texture Element Extraction And Texture Synthesis Method And Its Application

Texture, texture synthesis and their applications have been one of the hottest topics in the areas of Computer Graphics, Computer Vision, and Image Processing etc. They developed very well in recent years. This paper is focused on the research of the algorithm and application of texture synthesis, and has made a series of achievements.Patch-based sampling texture synthesis algorithm is a successful and widely applied algorithm compared to other 2D texture synthesis algorithms. The algorithm brings good synthesis effect from most textures, and can achieve real-time synthesis. The disadvantage of the algorithm is that we need to use our experience to choose free parameters such as patch size and seam size manually during the process of synthesis. The synthesis result will not remain the features of the input picture because of either too big or too small free parameters. In order to fix this problem, we analyze and summarize the characters of stochastic textures and structured textures. After lots of experiments, we find the characters of the possible minimum value of the patch size in the patch-based sampling texture synthesis. Furthermore, we find the two characteristics of the stochastic textures, and define the definition of the texels of the stochastic textures. According to the above achievement, we design two different ways to abstract texels from the two kinds of textures. As to the structured texels, we design the image segmentation based method to abstract texels because they have obvious boundaries. As to the stochastic texels, we abstract texels by using the size-varied window feature tracing based method. The experiment results prove that the two methods which are put forward in this paper bring good abstraction effect to both two kinds of textels. The paper also concludes two principles used to estimate free parameters according to texels size, and choose free parameters according to these two principles during the texture synthesis.Simulating traditional paintings using computer becomes one of the active research topics in Non-Photorealistic Computer Graphics. Research of simulating western paintings such as canvas and water color has already achieved visual effects. In recent years, scholars have achieved many successes in the area of Chinese Painting simulation. This paper present a Chinese ink painting simulation method based on photos. First, use region segmentation method to separate the foregroundand background of the photos, calculate and mark image, and then use patch-based sampling texture synthesis algorithm to synthesis the paper background, while using diffusion operator based ink diffusion effect simulation method to the foreground. Finally, merge the foreground and background according to the marked images. The experiment result has already turned out the Chinese ink painting conception, so it's an effective makeup to the current methods. This method is more applicable to people who has no painting background compared with the painter model based interactive system method.This paper is supported by the following projects sponsored by NWPU graduate student innovation fund: 'High Performance Texture Compression and Synthesis System based on Content', 'Theory and applicable algorithm of Fractal Video Frequency Compression with High Compression Radio' and 'Image Processing System Based on Fractal'.