SIA OpenIR  > 光电信息技术研究室
基于SIFT的稳像算法的硬件实现
Alternative TitleA hardware implementation of image stable based on SIFT
李博1,2
Department光电信息技术研究室
Thesis Advisor朱丹
ClassificationTP391.4
Keyword电子稳像 特征点检测 Sift 尺度空间
Call NumberTP391.4/L31/2011
Pages49页
Degree Discipline控制理论与控制工程
Degree Name硕士
2011-05-27
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract摄像机成像过程中由于拍摄装置与景物的相对运动造成图像序列的不稳定,稳像技术研究的目的就是从输入的视频图像序列中去除因为摄像机的随机运动而引入的图像扰动,使图像序列稳定。利用纯数字图像处理的方法进行图像稳定研究是当前稳像研究发展的趋势,即根据图像内容确定图像之间的运动向量。运动向量的获取大致分为基于灰度与基于特征两类,基于特征的方法使用图像中少量代表性的信息来估计运动参数,可以较好的提高匹配速度。基于特征点的图像局部特征可以有效应对局部遮挡、物体变形等局部信息的丢失。由于摄相机可能存在的旋转、平移、焦距变化等随机运动,检测出的特征点必须具有旋转、平移和尺度不变性。 本文主要研究电子稳像系统的实时实现,针对稳像对SIFT算法进行了加速改进,并详细给出了该算法在FPGA上实现方案。首先,详细讨论了基于尺度空间的特征点检测和基于生物视觉原理的特征点匹配方法来完成图像匹配的算法——SIFT算法。该算法可以良好的应对图像发生旋转、平移、尺度变化,对亮度变化、仿射变化、噪声也具有一定的适应能力。然后,给出了该算法针对实时实现的改进,并验证了结果的精确性。最后,本文详细给出了基于SIFT特征提取算法的稳像系统在FPGA上的实现,仿真结果表明,经过改进的算法可以达到实时的结果。
Other AbstractUnstable image sequences were acquired because of the relative movement between camera and scene. The purpose of image stabilization is to remove image disturbance because of the random motion of camera from input video. Trend of image stabilization is using pure digital image processing methods, that is obtaining motion vector from image content. There are many methods and could be divided into two class, based on features and based on gray. Methods that based on features estimate motion parameters using representative information of image can improve speed greatly. Local features based on feature points can treat lost of local information because of partial occlusion or object deformation. The features must be at least partially invariant to illumination, 3D projective transforms, and common object variations. This paper is mainly about real time implementation of electronic image stabilization, the improvement of SIFT and the implementation on FPGA. First, introduces SIFT feature point detection based on scale space and neurons in inferior temporal cortex. This method can treat invariant to illumination, 3D projective transforms, and common object variations. Second, presents improvement of the method for real time implementation. Last, presents detail of the parallel hardware architecture system on FPGA, and it gives a absolutely real time result.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/9230
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
李博. 基于SIFT的稳像算法的硬件实现[D]. 沈阳. 中国科学院沈阳自动化研究所,2011.
Files in This Item:
File Name/Size DocType Version Access License
基于SIFT 的稳像算法的硬件实现.pd(795KB) 开放获取LicenseApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李博]'s Articles
Baidu academic
Similar articles in Baidu academic
[李博]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李博]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.