基于GPU的医学图像快速体绘制算法  

Fast Volume Rendering of Medical Images on GPU

作　　者：卜祥磊[1] 冯前进[1] 秦安[1] 黄二亮[1] 陈武凡[1] 

机构地区：[1]南方医科大学生物医学工程学院,广东广州510515

出　　处：《中国医学物理学杂志》2009年第3期1167-1171,共5页Chinese Journal of Medical Physics

基　　金：国家重点基础研究发展规划(973)项目(No.2003CB716104)

摘　　要：目的:将传统的光线投射体绘制算法在具有可编程管线的图形处理器(GPU)上重新实现,将耗时的三线性插值和采样过程放在GPU上进行,提高重建速度。方法:首先将体数据和传递函数映射为纹理并将其载入到显存,接着通过对顶点着色程序和像素着色程序的编写将光线进入点、离开点的计算以及图像的合成运算移入GPU中,最后通过调整传递函数来实现不同的绘制效果。通过使用渲染到纹理技术,将绘制的中间结果保存到纹理,并以此来避免使用着色器的动态分支功能。结果:与传统的光线投射算法相比,本文算法可快速重建出质量较高的图像。结论:实验表明,在同等绘制质量的前提下,该方法的绘制速度显著提高,能够满足医学影像可视化的实时交互需求,具有较好的临床应用前景。Objective： In order to improve the performance of volume rendering,we implement the traditional ray-casting-based volume rendering algorithm on the programmable graphics processing unit （GPU）. By the algorithm, we put the time-cost tri-linear interpolation and resampling steps in GPU. Methods： First,the data sets and the transfer functions are transferred into the video as textures.Then the entry-point and the leave-point of the ray and the compose of the ray are calculated by the vertex shader and the pixel shader.Finally,different rendering effects are achieved by modifying the transfer fimction.The technique of rendering to the render target is used in our algorithm.We use textures to keep the intermediate results so that dynamic branch- ing can be avoided. Results： Our algorithm can get a much better performance compared with the traditional ray-casting-based volume rendering algorithm. Conclusions： The experiment shows that our algorithm is significantly fast and can meet the requirements of real-time interactive rendering on the premise of ensuring imaging quality.

关 键 词：体绘制 GPU 三线性插值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]