基于纹理硬件的大规模体数据快速绘制算法

Efficiently rendering large volume data using texture mapping hardware

针对大规模的体数据 ,有限的硬件纹理内存将大大降低算法的有效性 ,提出了一个新算法用于加速基于纹理硬件的大规模体数据的体绘制。基于一个新的 4级体数据装入流水线 ,算法在绘制前快速有效的对体数据进行预处理后 ,只将对最终结果图像有贡献的体元装入纹理内存并用于绘制 ,从而有效的降低了系统负载。同时 ,提出的算法支持对原始体数据分类阈值的交互修改与分类结果的快速预处理与交互显示。实验显示 ,与原始的基于纹理硬件的体绘制方法比较 ,本文提出的算法节省了 40 %到 6 0

Small volume data can be directly rendered with texture mapping hardware in real time. However, limited texture memory prevents the method from being used to render large volume data efficiently. In this paper, A new approach was proposed to accelerate volume rendering with texture mapping hardware. Based on a new volume loading pipeline, the volume data was preprocessed efficiently before rendering. Only the volume data that contain object voxels are loaded into texture memory and resampled for rendering. Test shows that about 40% to 60% of the rendering time is saved by our method for large volume data.