deblurring ratio is. As general, if we want to get a good de-
blurring effect,we must choose a smaller ratio of the reference
beam to the object beam,the modulations of the fringes are very
low, so the efficiency of the filter is small.' But'in © this new
System, the inverse filtering function can be got automatically
with the help of the mask. We only choose the ratio R/Ajaccord-
ing to the linear recording condition, so we can get higher mo-
dulation interference fringes.The other methods can not compare
with this method in efficiency. And it is verified by experi-
ments that there are big differences in efficiency if different
photographic condition is used. So it isveryv important to op-
timize the recording and processing conditions.
3°. The anti-noise property
The holographic system is damaged by various noise, espe-
cially the granularity noise due to random aggregations of the
silver grains in the emulsion. Heistrom developed an expression
for the optimum filter function, when random zero-mean,additive
noise is present. As the criterion for best filtering,the prin-
ciple of the least mean-squares is used in his derivation. The
optimum fiiter function is
Po / ® 2
Qo / D +1/ IH} 2
where ¢, , 9, are the power spectra of the noise distribution and
the image respectively. He showed that the smaller the 9,/9, is,
the stronger the noise is restrained. From part II we can give
an another form of our filter
2XfR / Ao 1
| *(OCf2R7/A/|C,H |? C,-H
here ^| 9$4/9«9« A4? f£? (R/Ao )? 40.0014 (for à 6328A, f - 600mm, R/As 1/10),
it is very small. So the filter we have got is optimal with
respect to the additive noise.
Another advantage of this system is that the output image
is composed of N pictures. So the SNR of N-channel filtering
system is increased by N times than that of a single one. The
affection of the system noise is decreased greatly.
t=H" (18)
(19)
Le =
V. Experimental
The linear motion blur image(Fig.2a) simulated in the labo-
ratory and the blurred remote-sensing image (Fig.2b) were pro-
cessed by this: system.The over all Gammas of two blurred images
were controlled to be 2 approximately. So that their amplitude
transmittances are proportional to the irradiation distribu-
tions of the blurred objects.
The image shown in Fig.2b was the deblurred one of Fig.2a
using the 3-channel inverse filtering system. The sharpness of
the restored picture was improved greatly and most of the de-
tails blurred was almost restored.
Fig.2 is the restored remote sensing image. In order to
evaluate the capacity of the filtering system, we chose a part
of the resolution test marks from an aero-photograph. From the
ratio of ground speed to the flight height (V/H) of the aero-
plane as well as other references the blur width was figured
out, which was 100um. If we observe this photo directly, the
42