ISPRS Commission II, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002 
  
A GLOBAL OPTIMAL REGISTRATION METHOD FOR 
SATELLITE REMOTE SENSING IMAGES 
Gongjian Wen *" *, Deren Li? ,Liangpei Zhang, Xiuxiao Yuan* 
a LIESMARS, Wuhan University, Wuhan, Hubei, China - dli@wtusm.edu.cn 
b ATR National Lab National University of Defense Technology, Changsha, Hunan, China - wengongjian(@sina.com 
Commission III, WG IIU/6 
KEY WORDS: Image Registration, Remote Sensing, Multi-temporal, Multi-sensor, Change Detection, Global, Simplex Method, 
Genetic Algorithm. 
ABSTRACT: 
One of the main obstacles in image registration is the precise estimation of a mapping function that determines geometric 
transformation between two image coordinate systems. For conventional image registration methods, their registration results are not 
the global optimal, and accuracy is low because only a few local control points are used for the estimation. In this paper, we develop 
a global optimal method in order to get a registration approach with high accuracy. In our method, an energy function that is directly 
related to the parameters of the mapping function is defined in the whole image. Thus, estimation of the global optimal mapping 
function can be solved through energy optimization. In defining the energy function, we choose a strength measure that is based on 
contour edge points. It is demonstrated that the strength measure is insensitive to image radiometric distortion. Therefore, our 
method is applicable for various kinds of images, even for different sensors images. In order to solve the energy optimization, we 
design a pipelining hybrid framework that combines genetic algorithms (GAs) and a simplex method (SM). The GAs are applied 
firstly to look for a few initial guesses from some sub-images, and then the SM is employed to get the optima of the energy function 
near these initial guesses. It is found that the pipelining hybrid framework is not trapped in a local optimum, and converges fast. 
Hence, one of the advantages of our algorithm is that it successfully avoids advanced feature extraction and feature matching in the 
image registration. Its characteristics are of automatic and robust. Experimental results have shown that our method can provide 
better accuracy than the manual registration. 
1. INTRODUCTION There are mainly two classes of automated registration: the 
area-based and feature-based methods. In the area-based 
  
Image registration is a process of matching two images so that 
corresponding coordinate points in the two images correspond 
to the same geographic area. Image registration between two 
remote sensing images is a very important image pre-process 
step for data fusion and change detection [1], and its accuracy 
has a key impact on their post-process [2]. 
Existing image registration techniques are generally divided 
into three broad categories: manual registration, semiautomatic 
registration, and automatic registration. Generally they all 
follow three steps to register two images: firstly, a number of 
control points are chosen or extracted from the two images, and 
then these points are used to determine a mapping function. 
Finally the mapping function is utilized to resample the second 
image so as to bring it into alignment with the first image. 
Therefore, the precision of image registration is controlled by 
accuracy and veracity of the control points. 
In the manual registration, a large number of control points that 
are uniformly distributed in the whole image must be selected 
manually. It is a very tedious and repetitive task especially 
when the image size is very large. Therefore, it is necessary to 
introduce automated techniques so that little or no operator 
supervision is required. 
  
* Gongjian Wen. 
China , 430079. 
A - 394 
methods [3], a small window of points in the first image is 
statistically compared with the same sized window in the 
second image. The centres of the matched windows are the 
control points. Feature-based methods usually consist of two 
steps: firstly, the common structural features [4] are extracted 
from the two images respectively, and then the matched 
features are utilized to acquire control points. 
Almost all image registration techniques implement such a 
strategy in which a few local control points are exploited so as 
to determine the global mapping function. But accuracy and 
veracity of the control points are limited in real cases. Even 
though the control points are manually matched correctly, their 
measure accuracy is still on a pixel-level. Consequently, results 
from these techniques are not the global optimal, and the 
accuracy is not too high because a few points are not precise 
enough to introduce the global parameters. Therefore, it is 
necessary to estimate the mapping function in the global range. 
So far, few studies have been conducted on the global optimal 
solution though it is so important for precise image registration. 
Therefore, in this paper, our aim is to propose a global optimal 
image registration method in order to achieve a better 
performance of the image registration. 
Tel: 0862787211051, E-mail:wengongjian@sina.com, Add: LIESMARS, Wuhan University, Wuhan, Hubei,