Research On 2D Texture Synthesis And Its Applications

With the fast development of computer techniques, multimedia products are becoming more and more common in people's daily life. As for its growing importance, more and more new and advanced techniques are in need, especially in the fields of photo processing and virtual scene reconstruction. All of the above set great demand for the development of computer graphics, image processing, virtual reality, computer vision and photo-realistic rendering. Texture is used to represent the rich details on the surface of the objects, so it can be also used to depict the natural phenomenon that has repeated features. As for the synthesized image and animation, texture shows extraordinary importance. Due to all of the above, texture techniques are drawing more and more extensive attention, and enjoy promising development.This paper mainly focuses on the algorithm and application of two-dimensional texture synthesis. We present a new texture synthesis algorithm which is called Patch-Based L-shaped Neighborhood Search (PBLNS). PBLNS is simple and fast, and also it shows great synthesis results on natural textures and half-structured textures. In the selection phase of texture synthesis, we put forward a patch-based L-shaped neighborhood search algorithm, which is quite different from the past pixel-based algorithms. Its patch-based neighborhood makes full use of its distinct L-shaped neighborhood to reduce the search space, as well as preserve the structure continuity and integrity of the texture. In the quilting phase of texture synthesis, we calculate the minimum error boundary of the overlapped region through dynamical programming, as well as adopt the feathering to further smooth the boundary.Based on the extension of PBLNS algorithm, we also present a new constrained texture synthesis algorithm, which extends the L-shaped neighborhood to square shape. In this way, we can further include the right and bottom side of the square-shape neighborhood to make better integration of the selected patch and the square-shape neighborhood texture. And all of the efforts finally resolve the discontinuity problem of right and bottom boundary during constrained texture synthesis. We also present a few application examples of this kind.Besides, this paper presents new multi-source texture synthesis algorithms. This algorithm can enable user to adopt different texture characteristics for matching and selection based on different applications, as well as improve the flexibility and applicability of multi-source texture synthesis, which isn't limited within color matching. What's more, itextends the mixed weight of multi-source from one-dimension to two-dimension to facilitate the output texture layout of each sample's texture characteristics. Also, several application examples are given.