基于矢量量化压缩的大规模体数据直接绘制.pdfVIP

第 35 卷第 3 期 Vol.35 No.3 t 高大学学报〈自然科学版〉 2∞5 年5 月 JOURNAL OF SOUTHEAST UNIVERSITY (Natural Science Edition) May 2005 基于矢量量化压缩的大规模体数据直接绘制 吴仲乐 鲍旭东 罗立或 (东南大学生物科学与医学工程系，南京 21∞96) 摘要:针对大提模体数据通常不适合于凄件三维纹理加速绘制的问题，使用基于这域分裂的 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.并在望 彭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荒模体数据的局部感兴趣区域的体绘制.实写金结果表堤，在不同压 缩比和信噪比下，大规模体数据绘梧的图镶仍然具有较高的质量且维持高鞍速率，绘制性能满足 实时交互. 关键词:体绘羁;矢量量住;体数据压缩;三维纹理 中医分类号: TP391 文献标识码:A 文章编号: 1∞1 -0505(2005)03心475-05 Direct rendering of large-scale volume data based on vector quantization compression Wu Zhongle Bao Xudong Luo Limin ( Departrnent of Biological Science and Medical Engiu出ring ， Sou自由st Universi号， Nanjing 21∞9岳， China) Abstract: Large volume data usually cannot be adapted to 3-D texture for hardware accelerated rendering.τo compress the volume data to the texture size ，姐 LBG algorithm of vector quantization based on region split me血。d is used to make code of compressed of large scale volume data wi注I high compression ratio. The prograrnmable shader of graphics processing unit decodes the 3-D texture of compressed volume data and renders the decoded data in real幽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.τhe rendering results show high frame ratio under different compression ratios and signal 10 noise ratios (SNR) with good visual quality , so 出e performance can support real-time interaction. Key words: volume rendering; vector quantization; volume compression; 3-D texture 体绘制是科