实时图像处理及其在医学中的应用

【作者】 伏思华；

【导师】 雷志辉；

【作者基本信息】 国防科学技术大学 ， 人机与环境工程， 2002， 硕士

【摘要】 本文提出了一种利用高性能图像采集卡在不使用DSP的情况下实现实时图像处理的方法。并将该技术应用于医学领域，对培养瓶中生长的心肌细胞的搏动情况进行实时记录，研制出一套心肌细胞搏动情况自动分析处理系统。主要工作包括以下几个方面： 1、详细论述了在WINDOWS操作系统下开发实时系统需要掌握的有关技术。介绍了图像采集卡和CCD摄像机的工作原理。结合MeteorⅡ和Domino-Mini两块图像卡的特性详尽说明了用它们实现实时图像处理的具体步骤。并对用高性能图像采集卡实现实时图像处理进行了经验总结。 2、简单介绍了动态图像处理的意义和一般方法。在此基础上介绍了如何用减影法进行图像处理，并对它进行改进提出了检测心肌细胞序列图像运动特征的一种方法。给出了实验数据曲线和结果分析，并对这种方法的优缺点进行了总结。 3、介绍了序列图像中相关跟踪方法的原理，改进了原有快速相关算法并用最小二乘曲面拟合的方法来得到亚像素的定位精度。用相关方法对心肌细胞序列图像进行处理，给出了实验数据曲线和结果分析。 4、研究光流法在动态图像处理中的应用，给出初步结果。以Visual C++6.0为开发环境，开发出运行于Windows 2000／NT下，具有良好用户界面、可靠稳定的软件系统，并设计了一个封装图像采集卡功能的类。 实验表明，本系统完全可以达到图像处理实时性的要求，对心肌细胞序列图像的处理结果也是令人满意的。更多还原

【Abstract】 This thesis bring forward a method to realize real-time image processing without using DSP but using high quality image grabber instead. Based on this technology, an automatic analysis and processing system, which can take the real-time record of the pulsation of myocardial cells growing in the culture flask, is designed and has been applied in the medical field already.The thesis is divided into 4 sections: Section I gives a detailed examination of the techniques relevant to real-time system developing under WINDOWS. This section also includes the operating principles of image grabber and CCD camera, an elaborate explanation of the concrete procedures of real-time image processing realized by image grabber Meteor I] and Domino-Mini, and an empirical summary on the real-time image processing realized by these high-quality image grabber. Section II attempts to introduce the significance of dynamic image processing and the oft-adopted methods. We focus our attention on how to process image with difference and bring forward a method analyzing the movement properties of the myocardial cells image sequences by the means of improving it. Finally, we provide an analysis of the experimental data with the pros and cons report followed. Section III begins with the principles of correlation tracking in image sequences, and then improves former swift correlation algorithm and achieves location precision of sub-pixel by using least-square-surface-fit method. Thus we manage myocardial cells image sequences with correlation and finally offer an analysis of the experimental data and the conclusion as well. Section IV studies the applications of optical flow in dynamic image processing and demonstrates the results in first stage. In this section, we also illustrate a system with a friendly users interface we developed with Visual C++ 6.0, which proves to be capable of running stably under Windows 2000/NT. A class encapsulating functions of image grabber is designed, too.Experiments show that the system can well meet the real-time requirements of image processing and the results of the processing of the image of myocardial cells order are also satisfactory.更多还原