Real-time Rendering And Fast Simplification For Large 3D Mesh Models

With its rapid development, 3D modeling technology is widely used in medical image, heritage, film production, 3D games and other fields. However, in the condition of current hardware, it is difficult to deal with the accurate and large models acquired by 3D laser scanners. Thus, many researchers focus on how to process these models, especially how to render them in real-time and simplify them fast. In the paper, we present the state of arts of the two problems and propose two algorithms to solve them.As for large model real-time rendering, we propose an algorithm, which is simple but high efficient. The algorithm first re-organizes the model into a tree structure. The upper nodes in the tree represents model with larger region but lower density; the lower nodes represents model with smaller region but higher density. Given a viewpoint for rendering, we choose only those appropriate nodes according to the distance between nodes and the viewpoint. In this way, we can reduce the number of triangles needed for rendering, and achieve the goal of real-time.As for large model fast simplification, we propose an algorithm, which is fast and can fairly maintain the features of an original model. The algorithm re-samples the model by rendering the model on GPU, and reconstructs the simplified model fast. Our method based on rendering can only reconstruct the visible part on the model when rendering. Therefore, we first segment the model into parts, then we simplify each part respectively, and finally we sew up the simplified parts to form a complete simplified model.