78
379
E(n)
= 2.» (3a)
Var(n) - ge AL (3b)
The variance of I. is therefore:
B gaz
Var I, = 0 (11) = Si/L (4)
2.2 SAR Image Statistics
From the image pixel statistics derived in par. 2.1 the covariance of the
image data samples, identified by their row and column coordinates (i.p)
and j,q), is given by:
| 2
E{(I..-m)(I__- z EG, -
(am gom) e ESSE som (8)
where,
m = image mean value = E(S, ;)
and
1
E(I,. 1.3) = MSA S imo 6
1) PG i 1J TAE: 1j p? (8)
For uncorrelated speckle,
1 T ; :
——ifin.m zo rl)s(id-p)síj-
d Sn nag o rsen eciog
Hence with (6),
1 2 ; ;
E(lpd m ES. uS + = ESS; = - 7
Er a (7)
where EZ is the unspeckled image variance,
LE hak ae
ES} = mag (8)
The image variance is thus given by:
Eum) -g 2, g 2
1j S p
ml+0 2
here c 2 pen (9)
W p a
is the image variance due to speckle.
A commonly used model describing the statistical properties of an
unspeckled (i.e. target) image is given by a two dimensional first order
Gauss-Markov process of which the covariance matrix values are given by:
= = mi st. 7 A1?
sd 7 PHS TS 9 m
where
e. = image variance = £13: 2)
Bg = correlation coefficient between row neighbouring pixels
by © correlation coefficient between column neighbouring pixels.
The covariance matrix values of an image effected by uncorrelated speckle
is then given by: