Multi-scale analysis based space array images registration

摘要：

Image registration is the core of target measurement, image recognition and computer vision. An approach to registration of space array images is studied on the basis of characteristics of space array images and principle of coarse-to-fine.. Firstly, a method of rough registration is proposed by selecting feature points on epipolar determined by relative position relationship among CCD through multi-scale image analysis in Gaussian scale space. Secondly, each image is divided into irregular connected region according to morphological connectivity theory. Lastly, a way to accurate image registration is established between irregular connected regions of space array images by means of dynamically selecting templates at feature points and gradually spreading regions from feature points with new fuzzy nearness degrees. Simulated experiments show validity and feasibility of the proposed method.

关键词： image registration multi-scale analysis image segmentation computer vision

作者: Tianzhen DONG Tingquan DENG Qiang FU Jiashu DAI Jin XU

作者单位: Laboratory of Fuzzy Information Analysis and Intelligent Recognition, Harbin Engineering University, Laboratory of Fuzzy Information Analysis and Intelligent Recognition, Harbin Engineering University, Agricultural Electricity Bureau of Naiman District, Tongliao Inner Mongolia, 028300, China Laboratory of Fuzzy Information Analysis and Intelligent Recognition, Harbin ngineering University,

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 274-277

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）