A Feature-adaptive Subdivision Method for Real-time 3D Reconstruction of Repeated Topology Surfaces
详细信息 查看全文
- 作者：Jinhua Lin ; Yanjie Wang ; Honghai Sun
- 关键词：GPU ; Feature ; adaptive subdivision ; Hardware tessellation ; Rendering
- 刊名：3D Research
- 出版年：2017
- 出版时间：March 2017
- 年：2017
- 卷：8
- 期：1
- 全文大小：
- 刊物类别：Engineering
- 刊物主题：Signal,Image and Speech Processing; Computer Imaging, Vision, Pattern Recognition and Graphics; Optics, Lasers, Photonics, Optical Devices;
- 出版者：3D Display Research Center
- ISSN：2092-6731
- 卷排序：8
文摘
It’s well known that rendering for a large number of triangles with GPU hardware tessellation has made great progress. However, due to the fixed nature of GPU pipeline, many off-line methods that perform well can not meet the on-line requirements. In this paper, an optimized Feature-adaptive subdivision method is proposed, which is more suitable for reconstructing surfaces with repeated cusps or creases. An Octree primitive is established in irregular regions where there are the same sharp vertices or creases, this method can find the neighbor geometry information quickly. Because of having the same topology structure between Octree primitive and feature region, the Octree feature points can match the arbitrary vertices in feature region more precisely. In the meanwhile, the patches is re-encoded in the Octree primitive by using the breadth-first strategy, resulting in a meta-table which allows for real-time reconstruction by GPU hardware tessellation unit. There is only one feature region needed to be calculated under Octree primitive, other regions with the same repeated feature generate their own meta-table directly, the reconstruction time is saved greatly for this step. With regard to the meshes having a large number of repeated topology feature, our algorithm improves the subdivision time by 17.575% and increases the average frame drawing time by 0.2373 ms compared to the traditional FAS (Feature-adaptive Subdivision), at the same time the model can be reconstructed in a watertight manner.