Abstract: An efficient algorithm is proposed to overcome problems existing in rendering large scale volume data.The original data is packed into small bricks and empty bricks are removed to save texture space.The index texture is also created for efficient packed texture access during rendering.The packed-index data structure allows us uploading lager scale volume data in GPU RAM entirely.The locations of empty bricks and high density bricks are marked in index texture and valid cube representation is generated before rendering for combining acceleration algorithm such as early ray termination and empty-space skipping.The algorithm can render large scale volume data efficiently and get the good quality of the resulting image.

Key words: volume rendering, scientific visualization, texture compression

摘要： 针对大规模数据体绘制效率低下的问题，提出一种算法：对体数据进行纹理分块打包，移除空数据块，并创建数据块的索引数据，绘制时通过索引访问打包后的纹理实现大规模数据完全载入显存，同时在索引中标记空数据及高密度数据块的位置，绘制前生成其有效的立方体数据表达，结合早期光线终止与空域跳过等加速技术，有效地实现了大规模的体数据的实时绘制，同时保证了结果图像的质量。

关键词: 体绘制, 科学可视化, 纹理压缩