大规模数据快速体绘制方法的设计与实现

Design and implementation of fast volume rendering for large data sets

体绘制是科学计算可视化的重要手段,针对大规模数据场体绘制计算量大、效率低的问题,提出一种基于哈尔小波变换的三维规则网格数据场快速体绘制方法。采用整体描述、局部细化的策略对转换后的数据进行组织、管理和调度,实现了多尺度分析和快速随机存取,并有效解决了CPU在处理大规模数据集时纹理内存的局限;采用基于CUDA(计算统一设备架构)的GPU加速技术提高体绘制算法性能,很好的实现了大规模数据集的交互式可视化。

Volume rendering is an important way of visualization in scientific computation.A fast volume rendering method based on Haar wavelet transform solving the problems of efficiency is presented in visualizing 3D regular grid datasets.The strategy of describing the whole and detailing the part is applied to organize and schedule the transformed data, which not only realizes multi-scale analysis and fast random access but also solves the limitation of video memory forthe large datasets.The technology of GPU-accelerated volume rendering based on CUDA is applied to improve the performance of volume rendering and the interactive visualization of large datasets is realized.