Abstract: In this paper we propose a method to accelerate 3D medical imaging surface reconstruction,which integrates characteristics of programmable GPU GeForce 8800 and Marching Cubes algorithm.Isosurfaces in every cube are extracted by the geometry shader and then triangles are generated for rendering.Utilizing Phong’s illumination model,these primitives emitted by the geometry shader are rasterized on the fragment shader to display.The main implementations including both isosurface extraction and rending run entirely on GPU without intervention of CPU.On the premise of ensuring imaging quality,Experimental results indicate that Large CT image dataset（512×512×400） can be perfectly modeled in real-time on a general PC platform.

Key words: GPU, GeForce 8800, geometry shader, Marching Cubes, medical image, GLSL

摘要： 提出了一种医学图像快速面绘制的方法,该方法将新一代图形处理器GeForce 8800的特性和MC（Marching Cubes）算法相结合。利用几何着色器的数据批处理能力在每个立方体中提取等值面并生成三角形带;在片段着色器上采用Phong光照模型对生成三角形渲染显示。建模和显示过程均在GPU上完成,对CPU的依赖低。实验表明,在保证绘制效果的前提下,该方法可在通用PC平台上实现大小为512×512×400的CT数据的实时建模,有很好的实用价值。

关键词: GPU, GeForce 8800, 几何着色器, Marching Cubes（MC）, 医学图像, GLSL