Research on the Active Volume Rendering

1.高分辨率实时体绘制系统的研究（1）大规模非结构网格并行体绘制系统我们开发了一个针对非结构网格体数据的单元投影并行体绘制系统。针对这个系统，我们提出了一种改进的早期光线终止方案，以修剪不可见细胞的细胞投影。为了减轻由于扫描转换中的视图依赖性和早期光线终止的动态行为而导致的负载不平衡，我们还在该系统中实现了动态负载平衡机制。(2)一个基于通用图形硬件的并行体绘制系统我们实现了一个基于通用图形硬件的并行体绘制系统。为了实现高帧率的大数据量渲染，我们还实现了两种数据压缩方案：基于自适应采样技术的体数据压缩和减少通信时间的中间图像压缩。因此，我们可以实现真实的实时绘制512×512×1024体数据使用四种商品的图形硬件。(3)一个可视化加速器我们提出了一个可扩展的并行体绘制的硬件架构，并开发了一个可视化加速器原型板（VisA Pro）。2主动绘制系统的研究我们提出了一个虚拟实验实时仿真系统，实现了仿真结果的同步可视化。所提出的系统基本上属于服务器-客户端模型。然而，不像普通的服务器-客户端模型，所提出的系统的客户端部分负责感兴趣的区域（ROI）部分的模拟，并使用其模拟结果进行实时用户交互，同时通过在服务器上同时执行相同区域的模拟并在客户端上校正模拟结果来保持模拟的准确性。我们把这种模拟模型称为“模拟缓存”。"

1 Development of High-Resolution Real-Time Volume Rendering Systems(1) A Parallel Volume Rendering System for Large-Scale Unstructured-GridWe have developed a cell-projection parallel volume rendering system for unstructured grid volume data. For this system, we have proposed a modified early ray termination scheme to prune the cell-projection of invisible cell. In order to alleviate load imbalance due, to view-dependency in scan-conversion and the dynamic behavior of early-ray termination, we have also implemented dynamic load balancing mechanism into this system.(2) A Parallel Volume Rendering System using Commodity Graphics HardwareWe have implemented a parallel volume rendering systems using commodity graphics hardware. In order to render large volume data with high frame rate, we have also implemented two kinds of data compression schemes : volume data compression based on adaptively sampling technique and intermediate-image compression to reduce communication time. Consequently, we could realize real time rendering of 512×512×1024 volume data using four commodity graphics hardware.(3) A Visualization AcceleratorWe have proposed a hardware architecture for scalable parallel volume rendering and developed a visualization accelerator prototype board (VisA Pro).2 Research on the Active Rendering SystemsWe have proposed a real-time simulation system for virtual experiments that realize the simultaneous visualization of simulation results. The proposed system basically belongs to the server-client model. However, not like the ordinary server-client model, the client part of the proposed system has responsible for the simulation of a region-of-interest (ROI) part and uses its simulation results for real-time user interaction, while preserving the accuracy of simulation by concurrently performing the simulation for the same region at the server and correcting the simulation results on the client. We call this simulation model "simulation caching."