| Image-based modeling has been widely used in movies, video games, protection of cultural relics, earth science and many other fields due to its strong realism, low cost, high automation and other advantages. However, long modeling time and large resources consumption have always been the two problems that 3D modeling cannot get rid of. Texture mapping and mesh generation have become the two bottlenecks because they are very time-consuming and resource-consuming.This paper adopts divide and conquer strategy to improve the texture mapping modular and the mesh generation modular. For texture mapping, a partitioning algorithm based on plane geometry area partition has been used to partition large triangle mesh model into some small models. Then we can perform texture mapping on these models in parallel. There is a certain degree of overlap between adjacent models to avoid cracks. For mesh generation, a partitioning algorithm based on bounding box partition has been used to partition large dense point model into some small models, so we can perform mesh generation algorithm on these models in parallel. There is a certain degree of overlap between adjacent models to avoid border effect. After mesh generation algorithm performed on these models, a mesh merging algorithm is needed to merge new generated meshes into a single topologically correct mesh. This paper makes use of OpenMP to improve current mesh merging algorithm to accelerate merging process.The experimental results show that, texture mapping and mesh generation have been significantly accelerated due to divide and conquer strategy. And the modified merging algorithm has a great improvement in speed. A shortcoming of this paper is that the partitioning algorithm is not very fast, which is yet to be further studied. |
