基于GPU的四维医学图像动态快速体绘制  

A GPU-Based Fast Dynamic Volume Rendering Method for 4D Medical Images

作　　者：秦绪佳[1] 王建奇[1] 朱思达[1] 郑红波[1] 徐晓刚[2,3] 

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310032 [2]大连舰艇学院装备系统与自动化系,大连116018 [3]浙江大学CAD&CG国家重点实验室,杭州310058

出　　处：《计算机辅助设计与图形学学报》2011年第11期1789-1798,共10页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61075118;60975016);浙江省自然科学基金(Y1100880);浙江省科技计划(2009C31106);浙江大学CAD&CG国家重点实验室开放课题(A0906)

摘　　要：传统的三维医学图像重建技术无法满足四维医学图像动态重建的需求,而四维医学图像庞大的数据量使传统重建技术很难实现高性能实时绘制.基于以上需求,提出了一种四维医学图像动态快速体绘制方法.首先采用GPU强大的并行计算能力,提出一种基于GPU、利用CUDA技术实现的光线投射算法;然后分析了算法框架、体数据及计算结果的存储策略、线程分配和寻址策略,并给出了四维动态绘制的流程.该算法将每一条投射光线的计算分派到GPU的每一个线程中并行执行,在保证绘制质量不下降的情况下,相比传统的光线投射算法其性能得到较大程度的提高.实验结果表明,该算法可以实现较高的绘制性能和流畅的动态显示效果.Traditional 3D medical image reconstruction method cannot meet the needs of dynamic rendering for 4D medical image. Furthermore, the traditional method would have very low performance due to the large amount of 4D image data. Based on the above requirements, a high-speed dynamic volume rendering method for 4D medical image is proposed. A high-speed ray casting algorithm based on the power of parallel computing of GPU with the architecture of CUDA is presented. The framework of the algorithm, storage strategies of volume data and computing results, thread allocation and addressing strategy are analyzed. The process of 4D dynamic rendering is shown. This algorithm can assign the computing of per casting ray to each thread in the GPU parallel execution. Compared to traditional volume rendering algorithm, this algorithm provides great performance, and preserves the advantage of ray casting in visual quality. Experimental results show that high performance and smooth rendering of dynamic display can be obtained by the proposed algorithm.

关 键 词：四维医学图像 体绘制 光线投射 CUDA GPGPU 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]