导　　师： 郭文明

学科专业： 081202

授予学位： 硕士

作　　者： ;

机构地区： 北京邮电大学

摘　　要： 显微镜作为一种光学仪器广泛应用于材料检测、医学研究等行业。在显微镜下,由于视域范围的限制,依靠人工采集成百上千具有重叠范围的清晰图像,并将其精确拼接成大幅面图像几乎不可能；另外显微镜的景深有限,采集一幅完全清晰的图像也是一个难题；拼接而成的超高分辨率图像基于现有的计算机软件难以实现快速显示。 本文在此背景下,使用软件控制自动扫描台移动以获取目标场景不同位置的显微图像；通过控制显微镜调焦旋钮,利用清晰度评价算法来实现自动聚焦以获取目标场景的清晰图像；最后采用快速傅立叶变换/(Fast Fourier Transform/)算法将所有显微图像拼接成一幅超高分辨率的图像；利用金字塔模型对超高分辨率图像进行分层分块存储,实现在本系统下快速浏览。解决了大范围自动扫描试样、显微镜下难于聚焦、超大图像的拼接合成和显示等问题,并在工程实践中获得良好的应用。 本文首先基于显微图像的应用过程中遇到的问题进行分析,论述了显微图像拼接系统的应用背景。 其次,对该系统进行需求分析,提出基于自动扫描台实现扫描、利用清晰度算法实现自动聚焦、基于快速傅立叶变换的拼接算法的设计方案,并对系统总体架构和功能模块进行了规划和设计,描述各功能模块的作用。 再次,对显微镜下难于获取清晰图像的问题进行分析,通过评价图像清晰度以实现自动聚焦获取清晰图像或者通过对多幅具有不同清晰位置的图像进行叠合构建清晰图像。 随后,详细说明了扫描采集流程,并利用自动聚焦原理,实现试样表面全自动覆盖扫描。 重点阐述了基于快速傅立叶变换配准原理和算法,计算相邻图像的配准点；并采用加权平滑算法有效消除了拼接产生的接缝。 最后本文介绍了超大图像的快速显� As an optical instrument, microscope is widely used in material testing, medical research and other industries. However, with the limitation of its vision scope, it is almost impracticable to collect hundreds of clear images with overlapping ranges manually and then stitch them into a large format image precisely. With the restricted depth of focus, it is also a hard work to gather a clear picture as a whole. Moreover, the ultra-high resolution image based on the existing computer software hardly achieves quick browsing. In this context, this thesis uses a software-controlled automatic scanning stage to acquire microscopic images from the target scene at different locations, obtains clear images by using target scene sharpness evaluation algorithm through controlling the microscope focusing knob. After that, the Fast Fourier Transform /(FFT/) algorithm is applied to stitch all microscopic images into an ultra-high resolution one; finally the pyramid model is employed for tiered and blocking storage of ultra-high resolution images to achieve quick browsing under the present system. As a result, issues such as auto-scanning samples in wide ranges, focusing under the microscope, mosaic and displaying of ultra-large images are resolved, which obtains excellent applications in engineering practice. The opening part presents the introduction. This part analyses the issues encountered in applications based on microscopic images, addresses the background of microscopic image stitching system. The second part goes on with the system requirement analysis, design proposals, planning and blueprint of overall system architecture and function modules, describing the role of each functional parts. Moreover, after an analysis on the difficulties in gaining clear images under microscope, the thesis raises methods and implementations to achieve clear images by auto-focusing through evaluating the image definition, or construct clear images through overlapping images with different clear locations. Su

关 键 词： 显微图像 自动聚焦 图像拼接 像素融合

领　　域： [自动化与计算机技术] [自动化与计算机技术]