GPU加速的表意式体模型线绘制算法  

GPU-based expressive line drawing algorithm for volumes

作　　者：张龙[1] 王毅刚[1] 吴向阳[2] 

机构地区：[1]杭州电子科技大学数字媒体与艺术设计学院,杭州310018 [2]杭州电子科技大学计算机学院.杭州310018

出　　处：《中国科学：信息科学》2014年第11期1385-1397,共13页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:60903085,61170150);浙江省自然科学基金(批准号:LY13F020048,Z1091077)资助项目

摘　　要：现有的体模型线绘制算法采用跟踪随机种子的方式得到特征线,存在严重的时空连续性问题.本文提出一种GPU加速的体模型线绘制算法,能够在每一帧抽取全部的特征线,从而彻底解决绘制中的不连续现象.我们首先提出一种并行的特征线抽取方法,在GPU的几何处理器中计算每个立方体单元中的特征线片段.为了解决可见性判定问题,我们采用splatting的方式生成深度图,取代已有方法中的光线投射方法,获得了明显的性能提升.采用自适应的深度偏移获得满意的可见性检测结果.为了将大量的体数据存储到高速显存中,我们通过剔除无关数据和压缩编码的方式,在不影响结果的前提下显著地减少了存储和带宽消耗.实验结果表明,本文算法能够生成时空连续的线绘制结果,并且绘制速度比已有CPU算法提高一个数量级.Existing line drawing algorithms for volumes extract feature lines by tracing randomly selected seeds, which is known to have severe temporal coherence problems. We propose a GPU-based line drawing algorithm framework for volume data. The algorithm extracts all feature lines in every frame, eliminating the discontinuity artifacts in successive frames. Firstly~ we propose a parallel algorithm to extract feature line segments in each cell in the geometry shader. To determine the visibility of feature lines, we use splatting to render a depth map for each frame, achieving much higher performance than the ray-casting method adopted in previous approaches. Correct visibility is obtained by using adaptive depth offsets. To store the large volume data in the high-speed video memory, we propose efficient data pruning and encoding schemes, which remarkably reduce the memory consumption and bandwidth, while maintaining the line drawing quality. Experimental results show that, our method can render high quality line drawings with temporal coherence, with a rendering speed that is an order of magnitude faster than previous CPU-based methods using seed-and-traversal.

关 键 词：绘制 体绘制 特征提取 图解 加速 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]