高实时性F-SIFT图像拼接算法  

High real-time F-SIFT image mosaic algorithm

作　　者：何宾[1] 陶丹[2] 彭勃[1] 

机构地区：[1]北京化工大学信息科学与技术学院,北京100029 [2]北京交通大学电子信息工程学院,北京100044

出　　处：《红外与激光工程》2013年第S02期440-444,共5页Infrared and Laser Engineering

基　　金：国家自然科学基金青年科学基金(61202431);高等学校博士学科点专项科研基金(20110009110001);中央高校基本科研业务费专项资金(2013JBM008);教育部留学回国人员科研启动基金

摘　　要：大场景视频拼接技术被广泛地应用于不同的领域,其应用要求精度高,实时性好。在对频域相关法、基于SIFT的图像拼接算法等方法进行深入研究的基础上,对其在配准精度、速度等方面进行改进与融合,提出一种具有高拼接性能、适应实时性拼接需求的F-SIFT图像拼接算法。该算法结合基于区域的图像配准方法和基于特征的SIFT两者优势,采用基于子图像块的频域相位相关算法,兼顾FFT的快速运算特点,提高图像拼接算法运算速度。实验对算法进行测试,验证F-SIFT算法的有效性。Large field video mosaic are widely used in different fields, it is required that mosaic is characteristic of high precision and good real- time. Based on deep researches of frequency- dependent method and SIFT image mosaic method, registration precision and speed is improved and merged. An F-SIFT image mosaic algorithm with high mosaicing and real- time performance was introduced. This algorithm incorporated the merits of the area- based image registration and feature- based SIFT, adopted phase correlation methods based on sub- blocks, which takes into account the characteristic of FFT quick calculation in order to speed the image mosaic algorithm. The effectiveness of algorithm was evaluated by the experimental results.

关 键 词：视频拼接 图像拼接 图像配准 尺度不变特征变换 傅里叶变换 

分 类 号：TN911.73[电子电信—通信与信息系统]