中文摘要：

采用多分辨率体绘制医学数据时,一般使用相同的阈值或者细节水平生成纹理,很难处理大规模数据,为此提出一种多分辨率纹理生成方法.首先采用基于方差加权香农熵的自适应分块细节水平选择算法建立原始体数据的统一划分多分辨率表示;然后采用分块纹理重组操作,生成具有更高压缩率的体数据多分辨率压缩纹理.文中方法已在GPU上实现,而且实验结果对比表明,该方法既能得到较好的体数据压缩率,又能完成高质量的绘制.

英文摘要：

To visualize medical volume data based on multi-resolution volume rendering,we often generate the textures with a uniform threshold or level of details（LOD）,but traditional methods are not efficient to deal with large-scale volumetric data.In this paper we propose a multi-resolution texture generation method that uses adaptive LOD,where the LOD is determined by Shannon entropy of the whole volume data with flat bricking.By using a bricking texture reorganization operation,we generate a compressed multi-representation texture.Our method is fully implemented on GPU,and the experimental results demonstrate that our approach with adaptive LOD can achieve overall higher rendering quality for a given memory constraint at a relatively high compression rate of the entire volumetric data.