【作者】 刘建军；

【导师】 肖德贵；

【作者基本信息】 湖南大学 ， 计算机应用技术， 2010， 硕士

【摘要】 体绘制是一种重要的三维数据场可视化技术,传递函数是体绘制过程中用以确定体数据与光学特征对应关系的关键步骤,对成像质量具有十分重要的作用。然而传递函数的设定往往需要大量尝试,困难且不直观,寻找好的传递函数已被列为体数据可视化中的十大难题之一。近年来,研究学者将智能算法引入到传递函数设计中,把寻找合适的传递函数的过程转化为参数优化问题,实现了传递函数的智能设计。但智能传递函数设计方法的绘制效果缺乏精确的客观评价,尤其对大型体数据的探索具有盲目性,并且自动设计效率较低。针对这些问题,本文从绘制评估和传递函数更新两个方面出发,对体绘制智能传递函数设计方法进行了研究。主要工作和成果如下：为了解决智能传递函数设计方法中图像质量评价方法各异、适用性差的问题,提出基于图像信息熵的图像质量评价方法。该方法利用图像信息熵度量图像纹理复杂度的特性,将其引入到智能设定的绘制评估模块,提高了智能传递函数应用于不同体数据可视化的效果。通过研究对比遗传算法、粒子群算法和量子粒子群算法,可以发现量子粒子群算法具有参数设置少、收敛速度快和全局寻优能力强的特点。因此,为了解决智能传递函数设定中自动设计效率低下的问题,提出基于量子粒子群的传递函数更新方法,通过其快速的全局寻优性能,达到了用户的绘制实时性要求。基于上述研究,结合图像信息熵的图像质量评价方法,提出一种基于量子粒子群算法的智能传递函数设计方法。该方法自动设计出适合不同数据场的传递函数。将该智能传递函数设定方法应到大规模地震数据可视化系统中,并在系统中为用户提供基于遗传算法、量子粒子群算法的手动调节与智能调节的传递函数设定工具。实验结果表明,该设计方法能够为地球物理学家在探索地震数据的过程中获取满意图像,能有效提高地震勘探的精确度和成功率。更多还原

【Abstract】 Volume rendering is a key technology for the visualization of three-dimensional dataset. Transfer function is used to specify the relationships between volume data and optical properties in the process of volume rendering. Specification of transfer function plays an important role in the rendering quality. Unfortunately, specifying a good transfer function is a difficult and tedious task, thus it has been referred as one of the top ten problems in volume visualization.Recently experts have been managed to introduce intelligent algorithms to specification of transfer function. They consider the design process of transfer function as a parameter optimization problem, and realize intelligent specification of transfer function. However there are still existing some problems such as poor rendering results evaluation methods, exploration blindness for the large volume datasets and inefficiency in the intelligent specification. This paper studies the intelligent specification of transfer function, especially on the image evaluation module and the transfer function updating module. The content of our paper is as follows:As for the problem existing in rendering image evaluation module, e.g. evaluation standards of image quality are different and used unpopular, we present an image evaluation method of image information entropy. Based on the comparison on types of methods of image quality, Image information entropy was introduced into rendering evaluation module, utilizing its intrinsic characteristic of measuring the complexity of the image texture. It improves visualization effect of intelligent transfer function implementing on different volume data.We also study and compare the Genetic Algorithm(GA), Particle Swarm Optimization(PSO) and Quantum-behaved Particle Swarm Optimization(QPSO) algorithms, further verify the characteristic of QPSO, e.g. less parameter, fast convergence and excellent global optimality. Aiming to solve the design inefficiency in intelligent specification of transfer function, we present an updating method for transfer function based on QPSO, utilizing the QPSO to update the iteration process. The result achieves the real-time requirement, which shows excellent global optimal abilities of QPSO. Based on the above mentioned, this paper presents an intelligent specification method for transfer function based on QPSO, which refers image information entropy as the image quality evaluation method. Our method designs kinds of transfer functions on different volume datasets. We implement the method in a large-scale seismic data visualization system. We also provide users with both manual and automatic transfer function design tools, which is based on the intelligent algorithms GA and QPSO. The experimental results show that our method can get the satisfactory image that the geophysicists’users desired in exploration of seismic data, thus improve the accuracy of seismic and success rate.更多还原