Study On Texture Synthesis And Replacement

Texture synthesis has a pivotal position in the field of image processing and texture replacement technology is one of hot research area in the computer graphics and computer vision. It has intimate relationship between texture synthesis and replacment technology, and it is widely used in the field of visual effect, game rendering, virtual reality, etc.After analyzing existing technology of texture synthesis and texture replacement, we study a new method of texture synthesis and replacement based on the texture statistical invariance and energy optimization. First, the texture statistical invariance is deduced and defined, and it is verified through experiments with a few kinds of typical texture image, including texture statistical invariance of average color and gradient. In order to guarantee the quality of texture synthesis in the process of texture replacement, the traditional energy optimization method is improved by proposing a new optimal search strategy based on nearest neighborhood search of Patch-match, which greatly improve the speed of optimization synthesis. Then the estimation method of texture illumination and deformation are studied based on the texture statistical invariance of the average color and average gradient respectively. Meanwhile an up-down sampling method is put forward to accelerate the estimate speed. Furthermore, to avoid the interference of image noise and jitter an adaptive sampling interpolation method based on TPS (Thin Plate Splines) is put forward for illumination and deformation estimation, which improve the estimate quality and further accelerate the estimate speed. Finally, we complete texture replacement through deformation field guided energy optimization texture synthesis method, combining with the recasting of illumination field. A lot of experiments has been conducted to verify the effectiveness of our method. Experiments results demonstrates very high visual observability and realistic of texture synthesis and replacement.