Vous êtes sur la page 1sur 2

In eqn. 12, p& = 0 as koa + m.

This provides the high frequency


solution given as
CkH?)(k&)

New time domain window


M. Szyper
Indexing terms: Discrete Fourier rransforms, Digitalfilters

rz Qk

(14)

Once c, is determined, it is possible to evaluate coefficients up and 6, using eqns 8 and 9.


Scattered field computations: With the formulation developed in the preceding Section, it is possible to evaluate the scattered field from a perfectly-conducting dihedral comer in the If-plane. If the scattered width, c ~ . of , corresponding two-dimensional dihe~ the dral comer reflector is known, the three-dimensional radar crosssection, qD, the truncated two-dimensional dihedral comer of reflector can be obtained by [3]
L73-D

A new window formula in the time domain is presented. It has mediate properties (figures of merit) into the Dolph-Chehyshev and Kaiser-Bessel windows, and its form is simple. The proposed window has four parameters, which allow its properties to he shaped for more applications.
Introduction: Time windows are used as weight functions for signals undergoing the DFT process. The characteristics of the windows described in the literature [l] are presented as a set of IO figures of merit (FOM) and determine their selection. However, these quantities are often in opposition to one another; for example, usually the lower the highest sidelobe level (HSL), the higher the equivalent noise bandwidth (ENBW). This makes selection of the windows difficult, because a choice of low HSL is beneficial in terms of the resolution of the low power spectrums elements, and a choice of low ENBW is beneficial in terms of the resolution of small averaging (smoothing) of the spectrum frequency. In this Letter the author proposes a new class of time window. Its characteristics allow a compromise to be made in the choice of window taking into account the resolution of the spectrums signals obtained in the DFT process. Window characteristics: The idea of the proposed window originates from the special properties of inverse hyperbolic functions, which create pairs of transforms with a similar shape [2]. The general, continuous and parametric form of the window can be described as follows:

= U2-D-

212

A0

where I is the longitudinal length of the three-dimensional object , and 1 is the free space wavelength.

21 0

-401. -80

-60 -40 -20

0
oi -9012

20

40

60

80

, I

w(t) = 1 - a (tanh b*&)

7_

(1)

Fig. 2 RCS against 9,-$./2 for four methods ($. = 98 9


~

_ _ - - UTD [I]

this Letter MM

In the presented formula w(t) represents the even function of the window for -T, 9 t 9 T,; fT, is the width of the window; a, b E R are the window parameters (real numbers); p, r E N represent the exponents of the window (natural numbers).
Table 1 Figures of merit of comparison windows :
~~

.. .. .. .. experiment

Hamming

To check the accuracy of the present method, our numerical data are compared with those obtained by MM, UTD [l] and the experiment in Fig. 2 which shows the angular behaviour of the backscattering cross-section when k = ko, a = I = 5.60881,f = 9.4GHz and 6 = 98. The number of modes used in the computa, tions is 101 and the angular behaviour of the present method is in excellent agreement with the results given in [l].
Conclusions: The mode matching technique is used to obtain an analytic series solution for scattering from a dihedral comer reflector in the If-plane. Some of the advantages of the present method are that it can be applied for arbitrary internal angles of the reflectors and its solution is also valid for relatively large scatterers with much faster convergence compared to that of the MM solution.
0 IEE 1995

FOM High sidelobe level [dB] Sidelobe fall-off[dBioct] Coherent gain Eauivalent noise IBW Hzl 3.0dB [BW Hz] Scalloping loss [dB]

window 43
4

New window 41.4


4

KeiserBessel window -69


4
0.40

0.54 1.36

0.46 1.56

1.80

3.26 6.0dB [BW Hz] Overlap corr.75 [%] Overlap car. 50 [/I
2.08

70.7 23.5

62.9 14.7

3.56 2.39 53.9

17 March I995

Electronics Letters Online No: 19950516

C.-G. Park, N.-H. Myung and S.-D. Choi (Department of Electrical Engineering, Korea Advanced Institute o Science and Technology f (KAZST), 373-1 Kusong-dong, Yusung-gu, Taejon 305-701. Korea)

References
and BALANIS, c.A.: Dihedral comer reflector backscatter using higher order reflections and diffractions, ZEEE Trans., 1987, AP-35, (ll), pp. 123S1247 2 HARRINGTON, R.F.: Field computation by moment methods (Macmillan, New York, 1968) 3 BALANIS, B.A.: Advanced engineering electromagnetics(Wiley, New York, 1989)
GRIESSER, T.,

The parametric form of the window allows the best shaping of its characteristics. Consequently, the choice of a, b and p , r is based on the desired FOM. Table 1 shows the FOM for the new window with the following parameters: a = 2.63, b = 0.178, p = 2, r = 4, 2T, = 1 s and sampling rate T, = 0.02s. To underline the value of the results, a comparison was made with similar data of typical windows: Hamming and Kaiser-Bessel (for a = 3). Figs. 1 and 2 present in graphical form the w(t) function and the positive part of its two-sided power spectral density function, calculated for a 1024-point FFT. The FOM values and Fig. 2 show that the proposed window is really a compromise between the reference windows with regard to HSL and ENBW. The position of the window in the Table, as quoted in [I], in the co-ordinate system: HSL against worst-case processing loss (W-CPL), seems satisfactory as well. Fig. 3 presents a fragment of this Table with new windows: w,(a=2.88, b=0.164), w,(a=2.63, b=0.178), w,(a=2.39, b=0.195), p = 2, r = 4 and w,(a=2.88, b=0.164), w5(a=2.63,b=0.178), w6(a=2.39, b=0.195), p = 2, r = 6.

ELECTRONICS LE7TRS 27th April 1995

Vol. 31

No. 9

707

-140 time,s

0
Fig. 1 New window in time domain

10

15

20

25

frequency,Hz
Fig. 4 Weak signal level detection by new window

Conclusions: The properties of the proposed window are comparable to the properties of Dolph-Chebyshev and Kaiser-Bessel windows. In terms of the windows currently employed, the choice of the described window makes it possible to obtain compromise FOMs. In other words, it is possible to obtain satisfactory resolution of the spectrum analysis in terms of the power frequency level of the analysed signals. Acknowledgment: This work was supported by the Polish Commit-

tee of Scientific Research, grant 8 S507 014 07.


0 IEE 1995 Electronics Letters Online No: I9950494

27 February I995

M. Szyper (Electrical, Automatics and Electronics Department, University of Mining and Metallurgy, AI. Mickiewicza, PL30-059 Krakow, Poland)

References
frequency, Hz Fig. 2 Power spectral density fmction o new window f
HARRIS, F.J.: Multirate FIR filters for interpolating and desampling, in ELLIOTT, D.F. (Ed.): Handbook of digital signal processing (Academic Press, San Diego, 1987) 2 BRACEWELL, R.: The Fourier transform and its applications (McCraw-Hill,Inc., 1965)

-50m

9-60Iu 0 n

?i -70-

Image enhancement with a fuzzy logic approach


J. Hsieh

* I 0 :

-801

Indexing terms: Image enhancement, Fuzzy set theory


~

- 9 0 1 , 1 , I I I I I 3.0 3.5 worst-case processing loss, dB

4-0
pg

An image enhancement technique is presented.The technique first segments the image based on a fuzzy logic function. A generalised median fdter (GMF) is then used for the noise suppression and

Fig. 3 Location o new window in co-ordinate system; HSL against Wf CPL

edge preservation in the membership function. The final image is derived with a set of nonlinear grey scale mapping techniques.
Introduction: Scout image generation is an integral part of the computed tomography (CT) scanning process. In most clinical applications, one or more scout images are generated prior to the start of a CT scan to provide anatomical landmarks for CT prescriptions. Scout images have also been used by doctors for the preparation of surgical procedures. For these applications, the contrast between the bony structures and the background is critical. Many attempts have been made in the past to enhance the visibility of these structures. These attempts, however, have met with limited success for two major reasons. First, the differentiation between the signals containing bony objects and their background is very small (less than 1% of the~dynamic range of the image). Secondly, the characteristics of the grey scale histogram distribution chances drasticallv from Datient to Datient, To overcome these difficulties, a dual energy approach was proposed [I]. This approach is based on the observation that the
Y~

0 Dolph-Chebyshev 0 new window A Kaiser-Bessel


Test of proposed window; To make at least a general evaluation of the efficacy of the window given by eqn. 1, Fig. 4 presents the DFT results of the complex signal x(t):

z ( t )= W ( t ) U ( t )
where

(2)

u(t)= sin20nt + 0.01 sin30nt + 0.001 sin42nt

(3)

The amplitudes and frequencies of the signals components were selected in such a way that the spectral resolution of the window on power level and frequency could be visually estimated without a special criterion, which seems sufficient.

708

ELECTRONICS LETTERS 27th April 1995

Vol. 31

No. 9

Vous aimerez peut-être aussi