【作者】 胡增海；

【导师】 张彩明；

【作者基本信息】 山东大学 ， 计算机应用技术， 2009， 硕士

【摘要】 本文主要基于CT医学断层数据来研究三维体绘制算法。三维体绘制算法是直接将体数据按一定的采样和映射规则合成,生成的图像具有更丰富的表现力。三维体绘制算法是当前研究的一个热点,被广泛应用于医学图像、CAD/CAM造型、工业设计、资源探测、虚拟现实等信息可视化领域。因此对三维可视化的研究,具有重要的学术意义和应用价值。Shear-Warp算法是被认为基于软件加速的最快的体绘制算法。本文主要对Shear-Warp算法的绘制质量和绘制速度两方面进行了深入的研究和探讨。在绘制图像质量方面,Shear-Warp算法为了取得较快的绘制速度而牺牲了图像质量。当用Shear-Warp算法进行绘制图像时,视线角度过大会使绘制生成的图像出现明显的波纹失真现象。主要原因是由于在重采样的过程中采样距离过大,违背了Nyquist采样定理,使得利用重采样的数据点无法恢复原来的数据信息。针对这个缺点,本文提出了一种消除Shear-Warp算法中波纹失真现象的新方法:基于插值逼近原理,先利用沿着主视方向相邻的四个原始采样点构造二次曲线;然后在构造的二次曲线上获得新采样点,再将新采样点添加相邻两层体切片之间来提高采样精度,达到消除波纹失真的目的。文中给出了加层原理、具体实现算法以及算法复杂度的分析。当然,在采用该方法改善绘制图像质量的同时,会影响到算法的绘制速度和消耗的空间。如何取得整体效果最佳的改进方法是我们以后要继续研究的问题。在绘制速度方面,Shear-Warp算法是最快的基于软件的体绘制算法,它对数据场和生成的图像采用一种巧妙的编码方式,在绘制的过程中,同时遍历数据场和图像,跳过数据场中透明的体素和图像中不透明的区域。这样大大加快了绘制速度。Shear-Warp是在单核CPU基础上,完全基于数据结构对算法的一种改进,与硬件完全无关。但是最近几年,随着多核CPU的普及,多核CPU在多任务处理和并行计算上体现出单核处理器无法比拟的优势,多核CPU逐渐取代单核CPU成为桌面和移动计算机的主流配置。因此针对多核CPU对体绘制算法进行速度优化十分必要。本文利用多线程、向量化运算等技术加速Shear-Warp算法,充分发挥多核CPU的特点,使其能够适应当前CPU的发展趋势,被更广泛的应用推广。本文利用CT医学断层数据对改进后的Shear-Warp算法进行了实验,实验结果表明,改进后的算法不论在图像质量方面还是绘制速度方面都取得了显著的改进。更多还原

【Abstract】 The main research of this thesis is on the Direct Volume Rendering (DVR) based on the CT volume data sets. According to the certain sampling and rendering rules, Direct Volume Rendering algorithm directly render the volume data to get an image. DVR is one of the most important techniques in the field of information visualization which is widely used in medical image, CAD/CAM prototyping, industrial design, oil and gas exploration, virtual reality and so on. So the research on DVR is very significant in academia and practice.Shear-warp algorithm has been recognized as the fastest software-based rendering method. This paper mainly does some research on two aspects of Shear-warp: the rendering quality and the rendering speed.In the rendering quality aspect, Shear-warp algorithm loss the quality of the image for improving rendering speed. The big view orientation angle will cause serious staircasing phenomenon in the final image when we render the volume data set by Shear-warp algorithm The primary cause is that the resampling distance is so long that violate the Nyquist theorem in the resampling process. So the resampling data are unable to restore the original data message. In this paper we brought a new method for removing the staircasing phenomenon. Based on the theory of interpolation and approximation, a quadratic curve is constructed from four ordinal original sampling points along the viewing orientation. A new resampling voxel is obtained from the quadratic curve. Then a new slice is constructed between two original neighboring slices so that the sampling precision can be improved. Based on the new sampling slices, staircasing phenomenon can be reduced in Shear-warp rendering. In the paper we will show the adding slices principle, the specific algorithm and the complexity analysis. Surely, what we do to improve the image quality will affect the rendering speed and consume more space. How to get the best effect of the algorithm is still our further research.In the rendering speed aspect, Shear-warp algorithm employed a clever volume and image encoding scheme, coupled with a simultaneous traversal of volume and image that skips opaque image regions and transparent voxels. This will speed up the rendering speed. Shear-warp is totally based on the data structures to do some improvements that have nothing to do with the hardware. But recently, as the multi-core processor is more and more popular, it becomes the main stream processor in common PC and notebook. The multi-core processor has incomparable advantages to single-core processor in the aspect of multi-processing and Parallel Computation. It’s very necessary for the volume rendering algorithm to be optimized based on multi-core CPU. In this paper we use multi-thread and vectorization methods to accelerate the Shear-warp algorithm which make full use of the multi-core CPU Characteristics. That makes it possible for the Shear-warp algorithm to become moreand more popular.In this paper, we do some experiments on the improved Shear-warp algorithmwith CT volume data sets. The experiment results show the efficiency of the improvements on both rendering quality and rendering speed.更多还原