【作者】 张华；

【导师】 李祖枢；

【作者基本信息】 重庆大学 ， 模式识别与智能系统， 2002， 硕士

【摘要】 虚拟环境中实时照相系统的研制是为了满足体验经济的发展对数字应用系统的需求。对其中关键技术环节----"实时人像提取"部分的深入研究，不仅能把虚拟现实技术（Virtual Reality Technology，简称VRT）的研究成果引入虚拟照像系统，使系统的性价比大大提高、"人性化"功能加强。同时，也对解决虚拟现实技术应用中所出现的"贵族化"、"计算繁琐"及"数据危机"等三大障碍,开发基于PC的虚拟现实应用系统，有一定的理论与实际的参考意义。本文从分析人类视觉特征入手，运用系统的观念综合平衡图像质量、实时性、与系统成本等技术指标，在对虚拟实时照相系统具体技术指标深入分析的基础上，提出了基于人类视觉系统（HVS）特性的实时人像提取算法。该算法依据人体视觉特征所决定的数字图象处理指标，参照人体视觉的信息处理机制，经过信息压缩、参数自适应的统计分类、形态学非线性滤波、图像融合、彩色图像增强等紧密配合的模块化处理过程，实现了鲁棒性较强的实时人像提取。算法中24位彩色图与二值特征图之间相互转换的过程,体现了利用变换域处理和信息分层处理的思想。文中对算法的各部分进行了原理与实现方法的深入分析，提供了在Matlab环境下的仿真图片。仿真结果表明，利用二值滤波处理较好地解决了传统的基于区域的图象分类中的边缘不连续的问题；利用边缘渐变方式实现的图像融合简洁有效；由灰度统计直方图均衡所实现的彩色增强处理降低了虚拟照相系统对环境照度的要求。原理分析、量化实验、系统实验、及部分算法在产品中的应用从多方面表明了论文所提方法的正确性。利用VC所提供的VFW（video for window）功能，在VC6.0的平台下实现了上述内容，应用表明，本文所提算法是可靠的。论文在虚拟环境中实时照相系统的硬件部分对系统结构、单片机操作单元部分进行了论述。更多还原

【Abstract】 The manufacture of real-time photographing system in virtual environment meets the needs that the development of experience economic towards digital application system. The lucubrating of its key technology part --"real-time portrait pick-up", has introduced the achievement of virtual reality technology into virtual photographing system, made the ratio of performance and price increase greatly, strengthened the "human" function. At the same time, to some extent it also has the practical and academic referenced sense in dealing with the three obstacles ,such as "noble"、"computer overloaded with details" and "data crisis", which appearance when virtual reality technology put into reality.This article begin with analyzing the character of human vision , making use of the notion of system to balance the several technology targets, such as image quality, real-time capability ,the cost of system and so on, and at the basis of thorough analysis toward the concrete technology target of the virtual real-time photographing system, putting forward the real-time portrait pick-up arithmetic based on human vision system. According as the digital image processing target determined by human vision character , through a series of processes ,such as information condensation, statistics classify base on parameter self-adaptation , morphological non-linearity filtering , image fusion , reinforcing of color picture and so on, this arithmetic realize real-time image pick-up with the development of robust. This arithmetic has also embody the idea of information layered dispose of diversification field dealing, which exist in the mutual transition between 24 bit color picture and two-value characteristic picture. In terms of individually part of this arithmetic, this article has analyzed deeply, and offered the simulation picture in the Matlab environment. The simulation result has indicated that using the method of two-value filter can solve the question perfectly, and the question is the edge discontinuity of traditional image classify base on region; The image fusion which make use of edge gradually change is sententious and efficient; The color image reinforcing which realized by grey statistics histogram equalization method has reduced the need of environment brightness in virtual photographing system. The application of principle analysis, quantum experiment ,system experiment, and some arithmetic in products has proved the exactness of the method we have provided from several aspects. Through utilizing the VFW(video for window) function which provided by VC system, has realized this arithmetic under the platform of VC6.0, the application has indicated that the scheme which bring forward in this article is exactness and credibility.<WP=6>In the part of hardware of the virtual environment real-time photographing system, this article has discussed several parts, such as system structure, single chip microcomputer assistant manage unit and so on.更多还原