144 
A METHOD ON HIGH-PRECISION RECTIFICATION AND REGISTRATION OF 
MULTI-SOURCE REMOTE SENSING IMAGERY 
Bin Liu \ Guo Zhang b , Xiaoyong Zhu c , Jianya Gong d 
a State key laboratory of Information Engineering in Surveying ,Mapping and Remote Sensing, Wuhan University, 
Wuhan, China, 430079,benjamin_lb@163.com 
b State key laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, 
Wuhan, China, 430079, guozhang@whu.edu.cn 
c Sehool of Remote Sensing and Information Engineering,Wuhan 
University, Wuhan,430079,China,zhuxytop@ 163.com 
d State key laboratory of Information Engineering in Surveying ,Mapping and Remote Sensing, Wuhan University, Wuhan, China, 
430079, jgong@lmars.whu.edu.cn 
KEY WORDS: Image registration; Feature detection; Mutual information; Hill climbing method; Tiny Facet Primitive 
ABSTRACT: 
For more and more applications on rectification and registration of Multi-source remote sensing imagery, studying in High-precision 
rectification and registration becomes significant. In this paper, we propose a method on High-precision rectification and registration 
of Multi-source remote sensing imagery, in which we use Mutual Information which is very general and has been used in many 
different image registration problems as similarity metrics. In order to enhance the search efficient, improved hill climbing method 
has been used. In addition, this method uses a High-precision algorithm: Tiny Facet Primitive. The Chinese-Brazil Earth Resources 
Satellite (CBERS) image and Landsat/ETM+ image have been used to test this method. In order to examine the result, we used a 
new method which combine the origin image and rectified image to be one image. The result shows very good. 
1. INTRODUCTION 
Image registration is the process of transforming the different 
sets of data which were acquired by sampling the same scene or 
object at the same or different times, by the same or different 
sensors, from the same or different viewpoint, from different 
perspectives into one coordinate system [1]. Registration is very 
useful in order to be able to compare or integrate the data 
obtained from different measurements. And it has wide 
application in remote sensing, medicine, cartography, computer 
vision, etc. 
With the increasing number of multiple platform remote sensing 
missions, many features may have a lot of different remote data 
at different spectral ranges, or different resolutions. The 
combination of the different data will allow for better analysis 
of various phenomena, as well as allow the validation of global 
low-resolution analysis by the use of local high-resolution data 
analysis [3]. 
Nowadays, applications on rectification and registration of 
Multi-source remote sensing imagery become more and more. 
LIU Shi-yin used registration of CBERS images and TM 
images to test glacier variations [8]. FAN Hui used combination 
of Landsat ETM+ and CBERS to detect the changes of 
Huanghe (Yellow) River [9].In this paper, we proposed a 
method on High-Precision rectification and registration of 
Multi-source Remote Sensing imagery, and use this on CBERS 
and Landsat ETM+. Experiment shows a good results on these 
two images which has different resolution and got from 
different time. 
2. IMAGE REGISTRATION METHODOLOGY 
In a whole, Image registration methods can be classified into 
two categories: feature based methods and intensity based 
methods. 
The method proposed in this paper, which is intensity based, 
consists of the following three steps: 
Fig.l. the process of registration 
2.1 Feature point detection 
In this step, we abstract feature points in the master image to be 
control points. Recently, a lot of different interest point 
detectors have been proposed with a wide range of definitions 
for what points in an image are interesting, in which the most 
popular operators are Moravec operator, Forstner operator. Due 
to their computational complexity, most of these algorithms are 
not well suited to large scale satellite remote sensing data 
processing applications [7]. In this paper, we used Moravec 
Operator to obtain a lot of feature points. The Moravec