G

■'m' m' l' u , 1<mX<mx, !<m'y<my, 1<l'<l x' y' '

ue {1,..., n}, denote the minimal gray level of data set u, and

1<mX<mx, 1<m' <my, 1<l'<l x> r u e {1,..., n}, the maximal gray level. Then each data set ue{1, ... , n} is rescaled according to c

r e {1,..., mx}, s e {1,..., my}, t e {1,..., l}. (41)

G e IR(mx,my ,l,n) here denotes the rescaled multispectral data set. An alternative approach would be rescaling according to the standard deviation of the gray-level distribution in each of the single data sets. This would reduce the effect of outliers on the rescaling procedure.

Rescaling enables eqivalent weighting of the single data sets in the subsequent vector quantization step; see Section 7.

0 0

Post a comment