Académique Documents
Professionnel Documents
Culture Documents
Three types of nonlinear transformations of the joint spectrum in nonlinear joint transform correlators
~NLJTCs! are investigated with the purpose of achieving the highest discrimination capability in target
location in a cluttered background: logarithmic transformation and the ~1yk!th law transformation in
combination with the limitation of the signal dynamic range and binarization by thresholding. By
computer simulation carried out on a set of test images, it is shown that application of these transfor-
mations in NLJTCs may considerably improve the correlators capacity to locate and recognize properly
small objects on a cluttered background, provided there is proper selection of nonlinearity parameters.
It is also shown that a moderate blur of the joint spectrum in such NLJTCs before nonlinear transfor-
mation is permissible, which simplifies the requirements of correlator optical alignment, the resolution
power of correlator electronic components, or both. 1997 Optical Society of America
Key words: Optical correlators, matched filters, joint transform correlators.
RO*~f!
Hopt~f! 5 , (1)
AVimsysAVxo@uB~f!u2#
where RO*~f! is the complex-conjugate Fourier spec-
trum of the reference ~target! object, uB~f!u2 is the
power spectrum of the background component of the
input image ~for object location! or average power
spectrum of objects to be rejected ~in object recogni-
tion!, AVimsys denotes averaging over the imaging-
system sensor noise, and AVx0 denotes averaging over
the unknown coordinates x0 of the target. The filter
described by Eq. ~1!, if realizable, is adaptive to the
background component of the image and therefore
provides the best possible performance for the given
individual input image. We refer to it as the OPAC.
The design of the OPAC requires that one know the
power spectrum of the background objects averaged
over the involved random factors. These data are
not known before the target is located and have to be
determined from the observed input image.
Fig. 1. Schematic diagram of the JTC with a nonlinear transfor- As a zero-order approximation of the background
mation of the joint spectrum. power spectrum one may use the squared absolute
value of the entire observed image spectrum uIM~f!u2:
ulation, a trade-off between the nonlinearity index k AVimsysAVx0uB~f!u2 < uIM~f!u2. (2)
and the nonlinearitys dynamic-range limitation
threshold. Being motivated by the results for the Approximation ~2! is based on the assumption that
NLJTCs with the ~1yk!th law nonlinearity and the the target object size is much smaller than the size of
limitation of the nonlinearitys dynamic range, in the entire image ~area of search!.
Section 5 we experimentally ~by computer simula- For a more accurate estimation of the background
tion! investigate BJTCs and show that, with an ap- power spectrum, one can use the following two mod-
propriately chosen binarization threshold, they may els for the representation of the image background
perform nearly as well as NLJTCs with logarithmic component:
and ~1yk!th law nonlinearities and that they are b~x! 5 w~x 2 x0!im~x!, (3)
rather tolerant to reasonable deviations of the bina-
rization threshold from its optimal value. In the b~x! 5 im~x! 2 ro~x 2 x0!, (4)
conclusion, we summarize the discussion.
where im~x! is the observed image, b~x! is its back-
2. Review of the Theory ground component, ro~x 2 x0! is the reference ~target!
We base our analysis on the theory presented in Ref. object, and w~x 2 x0! is a window function:
H
8. JTCs represent an attractive implementation of
0 within the target object
optical correlators. From the point of view of the w~x 2 x0! 5 , (5)
theory, optical correlators can be regarded as a spe- 1 elsewhere
cial case of localizationrecognition devices that con- For the model of Eq. ~3!, one can show that, with
sist of a linear filter followed by a unit for the the assumption of a uniform a priori distribution of
localization of the signal global maximum at the filter the object coordinates x0 over the picture area S,
output. It was shown in Ref. 8 that, for the problem
of target location in a cluttered background, the re- AVimsysAVx0uB~f!u2 < AVimsys@uIM~f!u2 ^ ~uW~f!u2yS!#, (6)
quirement for the highest discrimination capability of
such devices is practically equivalent to the require- where the symbol R denotes convolution and uWu2 is
ment that the filter provide the maximal ratio of the the squared magnitude of the window functions Fou-
signal value in the point of the output, where the rier spectrum.
target object is located, to the standard deviation of For the model of Eq. ~4!, one can show that, with
the signal in the background part of the output sig- the same assumption of a uniform distribution of the
nal. Since a number of random factors, such as sen- object coordinates x0 over the picture area S,
sor noise, an unknown reference-object position,
variability of the reference-object orientation, size, AVimsysAVx0uB~f!u2 < AVimsys@uIM~f!u2 1 uRO~f!u2#. (7)
H
assume that, for the majority of the spectral compo-
nents, x x # lim
LDR~ x! 5 . (14)
lim otherwise
uIM~f!u 1 uRO~f!u .. uIM~f!u z uRO~f!u.
2 2
(12)
The parameter lim defines the degree of the dynamic-
With this assumption, OUTNLJTC ~f! is approximately range limitation.
equal to An important practical issue in the design of the
JTCs of Fig. 1 is the required resolution power of the
OUTNLJTC~f! < log@uIM~f!u2 1 uRO~f!u2# TV camera that reads out the joint spectrum. It is
IM~f!* z RO~f! well known that the spacebandwidth product of an
1 optical system is usually much higher than that of
uIM~f!u2 1 uRO~f!u2
electronic imaging devices such as TV cameras.
IM~f! z RO*~f! This casts a restriction on the spacebandwidth prod-
1 . (13) uct of NLJTCs. On the other hand, the estimation
uIM~f!u2 1 uRO~f!u2
of the background-image power spectrum required
When a JTC configuration is utilized, the two last for the implementation of an OPAC according to ex-
terms displaying the correlation function are readily pression ~7! assumes that the input-image power
separated. The last term of expression ~13! resem- spectrum is smoothed by convolution with an appro-
bles expression ~10! when it corresponds to the OPAC priate window function, as was mentioned above.
with estimation of the background image component Experiments reported in Ref. 13 provide evidence
power spectrum by expression ~7! but without aver- that appropriate spectrum smoothing may provide
aging AVimsys. Therefore, one can conclude that the substantial improvement of the nonlinear correla-
nonlinear joint transform with the logarithmic non- tors discrimination capabilities. In the setup shown
linearity correlator can be regarded as an approxi- in Fig. 1, the limited resolution power of the TV cam-
mation to the OPAC and therefore promises an era ~as reported in Ref. 11! causes smoothing of the
coroutput
5 uIFT(CONV$LDR@log~uFT~corinput!u2!#, h~blr!%)z2.
(15)
and 5. The plots in Fig. 4 represent values of the In two of 16 test images no optimum was observed,
SNRV and SNRM for the logarithmic JTC as func- and the SNRM monotonically, although very slowly,
tions of the dynamic-range limitation parameter lim decreased with the increase of the blur parameter blr.
averaged over the set of test images. They show One can conclude from these data that the require-
that the logarithmic JTC is not very sensitive to the ment of the resolution power of the TV camera is not
limitations up to ~1027! of the entire dynamic range of very critical: The spatial bandwidth of the camera
the joint spectrum. may be 1.52 times less than that of the optics with-
Similarly averaged plots in Fig. 5 show how the out noticeable losses in the correlators discrimina-
discrimination capability of the logarithmic JTC de- tion capability.
pends on the blur of the joint spectrum. One can
observe the optimum degree of blur that tells that, 4. Nonlinear Joint Transform Correlators with the
although the expected gain in the correlators dis- ~1yk!th Law Nonlinearity
crimination capability derived from optimal joint- The distinctive feature of the logarithmic signal
spectrum smoothing is not high, the discrimination transform is that it compresses the signals dynamic
capability remains high over a rather broad range of range. Similar compression can be also achieved by
the degree of smoothing. Note that the optimum in the ~1yk!th law nonlinearity
the SNRM is less pronounced than that in the SNRV.
F~ . . . ! 5 ~ . . . !1yk, (16)
LDRh~x! 5 H0
1
x , lim
otherwise
. (18)
lim 5 1, . . . , 10. The results for the SNRM ~not
shown! are similar. These results show that From the experiments with the ~1yk!th law nonlin-
earity, one can conclude that even the simple
NLJTCs with the ~1yk!th law nonlinearity may dynamic-range limitation alone may substantially
have a considerably improved discrimination capabil- improve the NLJTC discrimination capability, pro-
ity in comparison with that of the JTC without non- vided the limitation threshold is properly chosen.
linear transformation of the joint spectrum ~case of This fact allows us to assume that BJTCs with the
k 5 1!. In terms of the SNRM, the observed aver- hard limitation according to Eq. ~18! may also have a
aged gain is exceeded 3 times. Comparison of the sufficiently high discrimination capability. The sim-
corresponding data for the NLJTCs with the ~1yk!th ulation results confirmed this conjecture. The sim-
law nonlinearity and the logarithmic JTC shows that, ulation was carried out with the same set of test
with an appropriate selection of the nonlinearity in- images according to the model:
dex k and the dynamic-range limitation threshold coroutput 5 uIFT(LDRh$@uFT~corinput!u2#%)u2. (19)
lim, the former performs slightly better.
As for the logarithmic JTC, the discrimination The simulation results are plotted in Fig. 8. The
capability of the NLJTC with the ~1yk!th law nonlin- graphs represent the SNRV for all 16 images of the
earity is not sensitive to the limitation threshold, test set as functions of the fraction of the joint-