146 
Fig.2. Traditional hill climbing method 
case of complete independence by means of the 
relative entropy or the Kullback-Leibler measure. 
2.2.4 Improved hill climbing method: In each pyramid- 
layer search process, we use improved hill climbing search 
strategy to reduce the number of calculation of correlation 
coefficient and guarantee the searched extremism point is the 
extremism point of the whole area. 
The traditional hill climbing method is an illuminating search 
strategy. By using illuminating information guidance, to let the 
search toward object and improved the efficiency of the 
resolution process. 
Nevertheless, due to its limitation, the hill climbing search is 
possible of obtaining the local extremism point. If we have 
illuminating information’s guidance, mountain-climbing 
method could solve the problem as soon as possible. However, 
if there is only one apex, the mountain-climbing method could 
promise to find the top, on the other hand, if there are many 
apexes, and the initial search point is in non apex area, the 
mountain-climbing would be very hard to get the real apex. 
Therefore, by improving the position of the initial point would 
help the hill climbing method get the optimizing apex point 
surely. In fig. 3 the improved search method make the search 
start from the initial place point position and follows the X 
direction that is POA to have a whole search. And find the 
whole apex in the X direction PI, and then follows P1B 
direction to have a whole search, to get the whole apex in the Y 
direction P2. This point has a very high possibility of be the 
main apex in this area. When search in X and Y direction, we 
could search with certain step length. It is unnecessary to 
calculate the correlation coefficients of every point, as long as 
the initial searching point is in the area of the main apex. Then, 
we can start search again in the original place by using 
traditional mountain-climbing method to calculate the main 
apex S2 quickly [12]. 
2.2.5 Eliminate the wrong match point: In this paper, we 
eliminate the match points which are larger than three RMS, by 
using two dimensional polynomial geometry relations of the 
generated corresponding points. 
2.3 Rectification and Registration based on TIN 
In this part, we first create the Triangulated Irregular Network 
Fig.3. Improved hill climbing method 
(TIN) by minimum distance method[10] in the two images. 
For each of the large number of triangles has three tie points 
Xi ,Yi ) ,( X' i ,Y' i ) ,i = 1 ,2 ,3 ,which can be used to 
calculate affine parameters: 
X' = aO + al X + a2 Y 
Y' = bO + bl X + b2 Y (3) 
Three points can get six equations, than we can calculate 
aO ,al ,a2 and bO ,bl ,b2 with which we can correct theAP' 
.IP' 2 P' 3 on the reference image to API P2 P3 on the 
master image. The process of rectification as follow: 
Fig.4. the process of resampling