The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part Bl. Beijing 2008 
—■— the first group 
the second group 
—*— the third group 
—*— the forth group 
—•— the fifth group 
—1— the sixth groi^p 
the seventh group 
——— the eighth group 
the ninth group 
Line displacement 
Figure 6. Relation of correlation to line displacement 
For the need to conduct small error curve fitting around the 
peak, the samples used should be in the peak vicinity. So, only 
five groups of correlation of line j and j-1 with displacement 
ranging from -2 to 2 are used in the curve fitting. The groups 
are fitted to a quadratic curve. The fitted equation and the 
desired displacement value are as follows. 
f i (xj) = a j *x/ 
d — x 
U j A max/,(*,.) 
bj'tj+Cj 
(4) 
To avoid curve fitting errors, the absolute value of acquired d,is 
constrained to be smaller than 1, conditioned correction is 
performed during estimation, the dj is substituted by 
displacement of the preceding line dj.j if the value is greater 
than 1. 
As the jitter is modelled to be horizontal shifting caused by 
deviated observation, the image can be restored by position 
correction of each pixel, thus horizontal shifting is performed 
according to dj, the idea is shown in Figure 7, for simplification, 
the linear pixel unmixing is adopted, each pixel is considered as 
linear combination of the original two adjacent pixels. 
nF V 
d J 
L 
Figure 7. Horizontal sampling 
the sampling equation is as follows 
Figure 8. Process of jitter removal 
After processing, jitter in the images is removed, shape of the 
edges are restored. The processed image regains geometric 
information of the objects and achieves better visual effect. The 
images before and after processing are shown in Figure 9. 
L{i, j) = dj * /(/, j) + (1 - dj ) * /(i +1, j) (5) 
The whole process of jitter removal is shown in Figure 8.