382
T
pee
Mead
C 75 log, 01 ; log, b(w,)dw,
where b(w ) is defined by:
b = E enc]
20, (1+r) F(w
2
)
2
2 2
)(1+r) F(wp) + L(1-0, )(1-0, )
£u
i
(1+04
2
m
r = ——
2
Is
The channel capacity for a given value of r (image mean square to
variance ratio ) as a function of image correlation coefficient (po =p
- 05) with the number of "looks" (L) as a parameter is shown in figure
la. :
From the parametric equations of (20) it can further be derived that,
C
a) the ratios e can be taken as a independent variable.
b) the channel capacity for an uncorrelated image is given by:
L ;
= nik a
Co=0 2 log, em. (bits/sample) (18)
The mutual information of an uncorrelated image for the chi-square speckle
distribution is derived in the annex, which results yield values which are
30 to 40% higher than those for the Gaussian channel. The notion that the
He Sauces monotonically to zero for completely correlated imagery
; provides a mean to estimate the exact amount of mutual
Ww ion Scis lines in fig. lai.
3.1 Information Content per unit area
The parameter which specifies the information content per unit image area
is particularly useful in comparing systems with different number of
looks. The information content per unit area is given by:
C uz C ee by, HL, (oy Jy, ] (19)
da dp doa dor IL, (Iva ry EL CE dd
where
das d, - spatial resolution in repectively azimuth and range
direction.