QQ客服 官方微博 反馈留言

题名	大规模显微图像拼接算法研究
作者	岳永娟
学位类别	工学博士
答辩日期	2006-06-03
授予单位	中国科学院研究生院
授予地点	中国科学院自动化研究所
导师	彭思龙
关键词	图像拼接 显微图像 拓扑图 回路一致 误差累积 多倍率 Image mosaicing Microscope images Topology graph Loop consistent Error accumulation Multi-power
其他题名	Large Scale IC Microscope Images Mosaicing
学位专业	控制理论与控制工程
中文摘要	由于摄像器材的局限性，对于较大的目标场景，普通的摄像机只能采集到其中的一个局部。为了扩大视场范围、获取目标物体的完整信息，需要扫描多幅边界上部分重叠的图像以覆盖整个目标物体，并把多幅重叠图像拼接成一幅高分辨率的完整图像。目前图像拼接已经成为图像处理、计算机视觉中的一个重要的研究课题，在摄影测量、虚拟现实、视频压缩、视频检索、增大视野范围、提高分辨率等方面有着重要的应用。 芯片显微图像拼接技术在集成电路反向设计中有重要的作用。芯片是一种具有多层结构的特殊对象。经显微镜放大采集后的图像，数量庞大且重叠区信息的同构现象突出，从而严重影响图像拼接的质量。考虑到这些因素的影响，本文基于图像的拓扑关系，提出了一种多层显微图像的拼接方法，可以避免这种局部无法对准的情形，从而实现多层显微图像的拼接。首先分别对同层和异层图像对准；其次利用3D拓扑图修正以消除空白区的影响；然后根据修正的3D拓扑图和多层图像投影坐标的误差建立多层图像的全局对准模型；利用非线性最优化方法求解具有全局一致性的多层图像拼接结果。 误差累积是拼接中的关键问题。误差累积的产生是因为图像序列构成回路的原因，基于此，本文还提出了一种基于回路一致的拼接算法，首先把图像的拓扑图分解为回路部分和非回路部分，然后利用回路一致关于旋转和位移的约束，对两部分分别进行优化。该算法可以消除误差的累积效应，提高拼接的质量。 针对芯片电路显微图像拼接过程中遇到 “孤立区域”而无法拼接的情形，提出了多倍率的拼接方法。通过采集不同倍率上的对应图像，利用小倍率上的图像信息估计大倍率上相邻图像的对准参数，在设定的对准误差条件下进行优化，可以得到较好的拼接解。最后，设计了整个拼接系统的工作流程，并给出了预处理环节的实现方法和实验结果。
英文摘要	Due to limitations of camera equipment, general cameras only capture a part for larger object scene. To extend the range of view and capture the whole scene, the camera may need to traverse the scene and scan multiple partially overlapping images and stitch them into one panoramic image of the scene. Recently, image mosaicing has been an active area of research in image processing and computer vision. More significant applications include photogrammetry measurement, virtual reality, video compression, video indexing, increasing the field of view and resolution of a camera. Chip microscope images mosaicing technology is important in the IC reverse design. Chip is a multilayer object. The number of chip microscope images is huge and the information is less or isomorphic in the overlapping region between two images, these influence the mosaicing quality seriously. Allowing for this, the algorithm based on the idea of topology graph can avoid the case that the local doesn’t align, a multilayer image mosaicing algorithm is proposed. The method first aligns images of same layers and images of different layers, and eliminates the effect of blank areas with 3D topology graph modification. Then it constructs the global image alignment model according to the modified 3D topology graph and the misalignment of multilayer image pairs. Last, it gets globally consistent estimation of multilayer image alignment parameters with nonlinear optimize method. Error accumulation is a key problem. It is due to loops of images sequence. Base on this, a new mosaicing algorithm based on loop consistent is proposed. The method first decomposes the topology into two parts, loops and non-loops. Then it separately optimizes the two parts with constrains about rotation and translation. This method can eliminate the effect of error accumulation and enhance the mosaicing quality. Mosaicing method based on multi-power images is proposed to avoid the case that the topology includes “isolated regions”. The method performs image mosaicing through estimating alignment parameters of high power images using corresponding low power images within the given alignment error range and then optimizing the mosaicing result. In the last, we design the whole mosaic system and give the method and result of preprocessing.
语种	中文
其他标识符	200318014603041
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/5930]
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	岳永娟. 大规模显微图像拼接算法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2006.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。