基于云端GPU的三维容积重建  

GPU-accelerated 3D volume rendering in cloud

作　　者：袁戎[1] 孙智[1] 谢庆国[1] 

机构地区：[1]华中科技大学生命科学与技术学院,武汉430074

出　　处：《计算机应用》2016年第A02期41-43,61,共4页journal of Computer Applications

基　　金：国家科技支撑计划项目(2012BAI13B06)

摘　　要：针对原有三维容积重建因计算资源需求大,导致硬件成本昂贵、难以应用于移动设备的问题,提出了一种基于云端GPU的三维容积重建系统。首先,在云端布置Web服务器和渲染服务器分别完成信息通信和图形计算的工作。接着,渲染服务器使用GPU并行计算重建过程中不同光线投射中的采样和Alpha混合,实现加速重建算法。最后,Web服务器和渲染服务器通过Redis内存数据库完成重建结果的传输,并通过Web服务器与客户端浏览器间的Web Socket连接实现交互与图像刷新。实验结果显示,不需安装任何插件,便携式笔记本和智能手机上的浏览器可以依靠云端GPU计算资源,在普通宽带的网络环境下对超过600 MB的体数据实现三维容积重建,平均交互帧率达到45帧/秒。Since the high performance computing of 3D volume rendering needs expensive hardware and is difficult to apply on mobile devices, a GPU-accelerated 3D volume rendering in the cloud was proposed. Firstly, Web server and rendering server were set in the cloud to manage the communication and graphics computing separately. Then, Rendering Server computed spacing and alpha maps from different ray-casting in parallel to accomplish GPU-aceelerated volume rendering. Lastly, in-memory data structure, Redis was used between Web server and rendering server while WebSoeket was used between Web server and client browser to implement interaction and rendered images reloading. Experimental results show that, based on the cloud GPU rendering, browsers on laptops and smart phones without any applets could implement 3D volume rendering to a dataset over 600 MB and the results achieved 45 frames per second averagely under a normal Web connection.

关 键 词：三维容积重建 云计算 GPU加速 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]