基于GPU加速的改进的光线投射算法研究

摘要

传统的医学图像体绘制算法大多通过CPU端实现,为解决传统算法存在绘制时间较长,交互不够流畅,且使用平台单一等问题,提出了一种基于图形处理器(GPU)的医学序列影像的实时体绘制技术.通过对不透明度值设置阈值,提前终止光线计算,并在光线遍历体数据时,调整采样间距,改进了光线投射算法.实验结果表明,对其中一组最大数据(931张切片)的绘制时间为1.3 s左右,交互时的绘制帧数在20~40之间.在绘制时间方面,不仅比传统的基于CPU的算法有大幅度提高,相比于前人的基于GPU的算法也有明显改进,加速比可以达到1.5左右.%The traditional volume rendering algorithms of medical image are mostly realized with CPU. In order to solve the problems of long rendering time,poor interaction and single platform existing in traditional algorithms,a real-time volume ren-dering technology of medical sequence image based on graphics process unit (GPU) is put forward. With the technology,the threshold of opacity value is set to terminate the light ray early,and the sampling interval is adjusted while the light ray travers-ing the volume data to improve the ray casting algorithm. The experimental results show that the volume rendering time of a group biggest data including 931 slices is about 1.3 s,and the rending frame is among 20~40 while interacting. The rendering time is faster than that of the traditional algorithm based on CPU,and the speed-up ratio can reach to about 1.5.