We have been expanding this experiment, and found that we are following
previous rob's post on cmap2rgb() etc.   Ok!, I got the problem right, I think.
The general solution, however, may not be so easy... Then, we'll try to solve
the problem under some local assumptions.

Kenji