Stationary in local meaning, Under this assumption Castleman (1979)
Proposed using the geometric mean filter. Here the power spectra of
the object image (and/or the noise) are functions of the position in
the image (position variant filtering),
5.2 Signal Dependent Noise
The restoration techniques presented in this paper were based on a
signal independent degradation model (eq. 4-1). The signal dependence
may be ignored for common restoration techniques, hut for high levels
of accuracy it must be taken into account, The signal dependent
property may be handled using homomorphic restoration techniques
(Stockham 1972, Oppenheim 1975), Consider the multiplicative noise de-
gradation model
g(x) = £(x) n(x) (5-1)
where the original image f(x) is degraded by a multiplicative noise
term n(x), Taking the logarithm of eq. (5-1) yields the additive
result
log(g(x)) = log(f(x)) + log(n(x)) (5-2)
Conventional linear filtering can now be used to estimate log(f(x)) and
by exponentiation we obtain the estimate of the original image f(x).
5.3 Conclusions
It is reasonable to assume, that the accuracy of matching of
stereoscopic aerial photos will increase with better image resolution.
Better resolution can be obtained using a suitable restoration
technique, The technique should be based On à priori knowledge of the
most significant degradation sources, This raises the question which
degradation sources are the limiting factors in the restoration process
of aerial photos. Three important degradation sources are:
- motion
- film grain noise
- atmosphere
There exist several nonrecursive and recursive methods to remove both
Space invariant and space variant motion blur (Sawchuk 1972, Frieden
1975, Rosenfeld and Kak 1976, Sondhi 1972, Gonzales and Wintz 1977),
Recently, however, new types of metric aerial cameras (Carl Zeiss Jena,
GDR and Zeiss Oberkochen, FRG) have been developed that have built-in
compensation for image motion in the flight direction. As a consequence
of this, longer exposure time and finer grained film can be used, thus
increasing spatial resolution, However, the problem of motion blur may
Still remain, the reason being that the aircraft with the equipment may
be subject to angular motion, the effect of which is in a direction
other than the flight direction.