基于矢量量化压缩的大规模体数据直接绘制  

Direct rendering of large-scale volume data based on vector quantization compression

作　　者：吴仲乐[1] 鲍旭东[1] 罗立民[1] 

机构地区：[1]东南大学生物科学与医学工程系,南京210096

出　　处：《东南大学学报（自然科学版）》2005年第3期475-479,共5页Journal of Southeast University：Natural Science Edition

摘　　要：针对大规模体数据通常不适合于硬件三维纹理加速绘制的问题,使用基于区域分裂的LBG算法对大规模体数据进行矢量量化压缩,使得其压缩编码适合硬件纹理空间大小.并在图形硬件的可编程渲染器中对压缩后的三维纹理体数据进行实时地矢量解码和绘制,在解码过程中通过设定掩码纹理给出了大规模体数据的局部感兴趣区域的体绘制.实验结果表明,在不同压缩比和信噪比下,大规模体数据绘制的图像仍然具有较高的质量且维持高帧速率,绘制性能满足实时交互.Large volume data usually cannot be adapted to 3-D texture for hardware accelerated rendering. To compress the volume data to the texture size, an LBG algorithm of vector quantization based on region split method is used to make code of compressed of large scale volume data with high compression ratio. The programmable shader of graphics processing unit decodes the 3-D texture of compressed volume data and renders the decoded data in real-time. Within the decoding process, a 3-D texture of mask volume data is selected to make the local rendering of region of interest (ROI) for very large scale volume data. The rendering results show high frame ratio under different compression ratios and signal to noise ratios (SNR) with good visual quality, so the performance can support real-time interaction.

关 键 词：体绘制 矢量量化 体数据压缩 三维纹理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]