Technology Research, Based On Energy Optimization Of Image And Video Texture Replacement

Texture replacement is a cross-research direction in the area of Compute Vision and Compute Graphics. It aims to solve the main problems of replacing the existing texture by a new one, which is based on preserving the shading and light effects as well as the original texture distortion of the concerned region.Texture replacement is widely used in many areas, such as image editing, TV program making, industry design, house decoration, virtual reality and so on. It satisfies people's demands of cultivating different atmosphere and textures on the condition of only one scene existing, thereby costs of production can be reduced and scene contents enriched.The thesis advances a novel texture replacement algorithm, which is based on energy optimization, and enables users to replace the textures in image and video. In hence, we can obtain more natural and realistic visual effects. The main contributions of this thesis are summarized as follows:1. An effective image texture replacement framework is proposed. We use intrinsic image to extract the shading image in the concerned region, so that the original shading can be preserved in the retexturing result, and the color of lights can be preserved too. Normal-From-Shading algorithm is also used to approximate the normal map of the concerned object, so as to preserve the distortion of original texture. In the last place, a method using Poisson optimization to compute the texture coordinates and then map the new texture to the concerned region is present.2. A novel and effective video texture replacement framework is presented. We propose a video texture replacement algorithm using the intrinsic video, which is extended from the intrinsic image algorithm. Experiment results show that our methods can preserve the shading and light effects of the concerned region in video sequence. By way of object tracking algorithm, optical flow algorithm, Poisson optimization and other techniques, we can get the correct texture coordinates of the concerned region, so as to prevent the texture drifting phenomenon. Consequently, users can replace the texture of video interactively.3. A new texture replacement method using mesh constraints is proposed. We use triangle mesh and rectangle mesh to enable the users to modify the direction and scale of the texture freely. The triangle mesh and rectangle mesh is used to guide the re-texturing process, and to handle the occusion problems between the video frames. Thus, the original perspective effects can be preserved and the result is more natural and more realistic.