P2-7-4 
Figure 3: Temporal Local Optimization Method. 
Figure 4: Multispectral Constraints Method. 
I « I % O'*«. 
I I I I I ll • • • 
lOpixels/frame 
Figure 5: Spatio-Temporal Local Optimization Method. 
lOpixels/frame 
Figure 6: Combined Method of Spatial Local 
Optimization and Multispectral Constraints. 
Figure 7: Combined Method of Temporal Local 
Optimization and Multispectral Constraints. 
Figure 8: Combined Method of Spatio-Temporal Local 
Optimization and Multispectral Constraints. 
...It • ly 
\ * * -k - 
• N » * f * «* - - — - 
-#******# 
** # . 0 - 
: : ; : 
lOpixels/frame 
Figure 9: Spatial Global Optimization Method. 
On the other hand, the results which were solved by 
temporal local optimization method (Figure 3) and 
combined method of temporal local optimization and 
multispectral constraints (Figure 7) were worse. It is said 
that temporal local optimization method performed well, 
only when the velocity of object is less than 1 pixel/frame. 
The result which was solved by multispectral constrains 
method (Figure 4) was also worse. Because brightness of 
channels (RGB) are similar, the solution is unstable (Figure 
10). 
In the result which was solved by spatial global 
optimization method, the optical flow tended to small 
(Figure 9). The magnitude of one of the estimated flow