MTF spectral-variation comparison of detector arrays used in

multispectral imaging systems by speckle patterns

Alicia Fernández-Oliveras, Antonio M. Pozo and Manuel Rubiño; Departamento de Óptica, Facultad de Ciencias, Universidad de
Granada; Granada, Spain

this proving indispensable in multispectral and colour-measuring

Abstract applications.
Currently, cameras based on CCD and CMOS detector Speckle is an interference phenomenon that occurs when
matrices offer excellent features in imaging systems if they are coherent radiation is scattered from a rough surface. Several
appropriately designed. Therefore, to investigate the suitability techniques can be used to generate the speckle pattern, such as
of the use of one or the other technology according to the different types of transmissive diffusers (ground glass [15],
specific application of the camera, the complete characterization fused silica [16], microlens arrays [17]) or integrating spheres
of the different types of detectors becomes necessary. [18-20].
In this work, we have analysed the quality of the images In the former case, an aperture situated in front of the
provided by different cameras by the speckle method. For this, integrating sphere enables us to specify the content of spatial
we have comparatively studied the Modulation Transfer frequencies of the speckle pattern. Two of the apertures used to
Function (MTF) at different wavelengths of the visible spectrum, date are the single-slit [18] and double-slit [19], both of which
for the detectors of a low-cost CCD video camera and of two present advantages and drawbacks [20]. In this work, we have
scientific cameras (a CCD and a CMOS). used a single-slit situated at the exit port of an integrating
For the CCD detector of the video camera, the highest sphere.
value of the MTF was reached at the lowest of the wavelengths It bears noting that in the works cited above, the systems
studied. Furthermore, the differences between the MTF curves analysed are based generally on scientific CCD cameras and
corresponding to the different wavelengths analysed become comparisons were not made between devices of different quality
more notable as the spatial frequency increases. or technology. Furthermore, neither was the MTF spectral-
In the case of scientific cameras, the behaviour of the MTF variation studied.
with wavelength does not present the same trend as that The aim of the present work is to apply the optical detector-
observed for the low-cost video camera. For each of the three characterization method, based on the measurement of the MTF
wavelengths studied, the CCD detector presented MTF values with speckle patterns, to the analysis at different wavelengths of
higher than those of the CMOS detector. the image quality provided by different cameras.
For this, we have comparatively studied the resulting MTF
Introduction curves at different wavelengths of the visible spectrum, for the
Cameras with imaging devices based on CCD and CMOS detectors of a low-cost CCD video camera and of two scientific
detector matrices [1] are being used more and more in such cameras (a CCD and a CMOS).
disparate fields of Science and Technology as Colorimetry,
Illumination, and Astrophysics. Currently, both types of devices Theoretical background
offer excellent features in imaging systems if they are The relationship between the theoretic power-spectral
appropriately designed. The consensus is that the two density known for a single-slit (PSDinput) and the measured
technologies complement each other and will coexist in the power-spectral density (PSDoutput) allows us to determine the
future, depending on the application involved [2-6]. MTF of the detector by means of the expression [18]:
Therefore, to investigate the suitability of the use of one or
the other technology according to the specific application of the PSDoutput ( ξ ,η ) = [MTF ( ξ ,η )]2 PSDinput ( ξ ,η ) (1)
camera, the complete characterization of the different types of where ξ and η are the spatial frequencies corresponding to the
detector matrices becomes necessary. horizontal and vertical directions x and y, respectively.
A system is optically characterized by the modulation PSDoutput is determined from the speckle pattern captured
transfer function (MTF), the determination of which enables the with the detector, being proportional to the squared magnitude of
image produced by the system to be evaluated from its response the Fourier transform of this speckle pattern. In the case of a
in spatial frequency [7, 8]. rectangular single-slit, PSDinput is given by [21, 22]:
For measuring the MTF of solid-state cameras, the (λz )2 tri⎛⎜ λz ξ ⎞⎟tri⎛⎜ λz η ⎞⎟⎤
2⎡
literature cites different methods that differ essentially in the PSDinput ( ξ ,η ) = I ⎢δ ( ξ ,η )+ ⎜ l ⎟ ⎜ l ⎟⎥
(2)
⎢ l1l2 ⎝ 1 ⎠ ⎝ 2 ⎠⎥
type of target or pattern used as the object. Thus, for example, ⎣ ⎦
methods use bar targets [9], random targets [10, 11], canted self- where tri(X)=1-|X| for |X|≤1 and zero elsewhere; < I > 2 is the
imaging targets[12], interferometric fringes[13, 14]. square of the average speckle irradiance; δ(ξ,η) is a delta
One of the methods to measure the MTF, established in our function; l1 and l2 are, respectively, horizontal and vertical
laboratory, is based on using a laser speckle pattern as the object dimensions of the single-slit; λ is the wavelength of the laser;
[15-20]. This method is suitable for analysing the detector and z is the distance between the single-slit aperture and the
independently of the camera lens, given that it does not require a detector.
lens to project the pattern. Furthermore, using a tunable laser
source, we can characterize the device at different wavelengths,

CGIV 2008 and MCS’08 Final Program and Proceedings 461

 Given the geometry of the single-slit, the PSDinput can be detector array is comprised of a matrix of 752x582 pixels
separated into frequencies ξ and η. The horizontal PSDinput(ξ,η) (horizontal x vertical). The horizontal spacing between centres
is the η=0 profile of PSDinput(ξ,η). This means the MTF can be of these pixels is 7.98 µm, providing a Nyquist frequency of
determined separately for x and y directions. In the present 62.66 cycles/mm in the horizontal direction by virtue of Eq. (4).
work, we determine the horizontal MTF. This can be done in a The scientific CCD camera had a high-resolution CCD B/N
similar way for the vertical direction. PixelFly array of 1360x1024 pixels with a centre-to-centre
spacing between them of 4.65 µm. Consequently, with Eq. (3)
Method taken into account, the Nyquist frequency of this detector is
107.53 cycles/mm in both directions.
Experimental Set-up The CMOS camera used was a CMOS B/N Atmos
Figure 1 presents the experimental set-up used. It is Areascan 1M30, the detector array of which had 1312x1024
composed of a tunable ion-argon laser source (130 mW), an pixels. In this case the pixel pitch was 5 µm in the horizontal as
integrating sphere to generate the speckle pattern (inner diameter well as in the vertical direction, corresponding to a Nyquist
of 152.4 mm), a polarizer to provide a linearly polarized laser- frequency of 100 cycles/mm given by Eq. (3).
speckle pattern, a single-slit (6 mm height and variable width), The width of the single-slit used was l1=1 mm in the case
and an optical bench to hold the detector, which is connected to of the low-cost video camera CCD detector, and l1=3 mm for the
the control card installed in a personal computer. two scientific detectors (the CCD and the CMOS).
The laser radiation is aimed at the entrance port of the Taking into account single-slit’s width and Nyquist
integrating sphere, generating the speckle pattern at the exit port. frequency, we can calculate the distance z between the detector
The aperture situated at the exit port of the sphere (single-slit) and the aperture by using Eq. (3) for each wavelength studied.
determines the content in spatial frequency of the pattern The corresponding values are collected in the following tables
registered in the detector. In these conditions, the linear for each detector analysed.
polarizer ensures that the PSDinput is given by Eq. (2) [22].
Table 1: Distance between the detector and the single-slit
aperture for the low-cost video camera CCD detector
Wavelength (nm) Distance detector-aperture (mm)

514 31

502 32

488 33

477 33
Figure 1. Experimental set-up for the measurement of the MTF of the
detectors. The aperture A (single-slit) and the polarizer P are situated at 454 35
the exit port of the integrating sphere.

With the single-slit, the MTF can be determined from a Table 2: Distance between the detector and the single-slit
single measurement without the need to move the detector, but it aperture for scientific camera CCD detector
must be situated at a distance from the aperture in such a way Wavelength (nm) Distance detector-aperture (mm)
that the maximum input spatial frequency is equal to the Nyquist
frequency of the detector [16, 18]. In this way, the MTF can be
514 54
determined in the largest possible frequency range, and thus
aliasing is avoided.
The distance z between the detector and the single-slit 488 57
aperture can be calculated by the expression:
l 457 61
z= 1 (3)
λ ξ Ny
where l1 is the slit width, λ the wavelength of the laser, and ξNy
is the Nyquist spatial frequency of the detector in the horizontal Table 3: Distance between the detector and the single-slit
direction. For a detector array with a centre-to-centre spacing aperture for scientific camera CMOS detector
between the photoelements ∆x, the Nyquist frequency is given Wavelength (nm) Distance detector-aperture (mm)
by:
1 514 58
ξ Ny = (4)
2∆x
In this work, measurements were made using the detectors 488 61
of three different cameras: a low-cost CCD video camera and
two scientific cameras (a CCD and a CMOS). 457 66
The video camera was a CCD B/N Center HICB347H,
connected to a Pinnacle Studio MovieBox DV control card. Its

462 ©2008 Society for Imaging Science and Technology

Data Processing experimental values of the horizontal MTF of each detector after
Once the detector was set at the corresponding distance of normalization at zero spatial frequency. For greater clarity, all
single-slit aperture, as indicated in the previous section, the the points corresponding to the 512 MTF experimental values
PSDoutput(ξ) was determined in the following way: are not shown.
For a given digitized frame of speckle data, a region of
500x500 pixels was selected. Each horizontal row of data is a
single observation of an ergodic random process. A fast Fourier
transform (FFT), which is a discrete Fourier transform, was
performed on each row of speckle data. The magnitude squared
in one dimension provided a single estimate of the one-
dimensional power spectrum, PSDoutput(ξ). These 600 spectra
were averaged, for a better signal-to-noise ratio in the
PSDoutput(ξ) [23]. To reduce the noise even further, the average
was taken for 10 frames.
The frames were stored in tiff format without compression,
using an integration time of 0.050 s for the scientific CCD
detector and 0.004 s for the CMOS detector. In the case of the
video camera CCD detector, the frames were extracted in tiff
format from a video recording captured with a rate of 25
frames/s for 1 s.
When a FFT is performed on a data set of length N, the
Figure 2. MTF experimental values of the low-cost video camera CCD
Nyquist frequency appears at the N/2 component of the FFT
detector at different wavelengths of the visible spectrum.
output. A ratio can be formed to evaluate the spatial frequency
ξn that corresponds to the n’th component as [20]:
ξ Ny ξ n
= (5)
N 2 n
Eq. (5) associates frequencies between zero and the Nyquist
frequency with FFT components from 0 to the N/2 component.
In this work we used N=1024, thus the total number of spatial
frequencies contained in the range from 0 to the Nyquist
frequency of the detector was 512.
Before processing, each digitized frame of speckle data was
corrected in order to reduce effects from the spatial noise of the
detector itself.
With respect to the spatial noise of a CCD, a distinction can
be made between the fixed pattern noise (FPN) and the
photoresponse non-uniformity (PRNU). The FPN refers to the
pixel-to-pixel variation that occurs when the array is in the dark,
and thus it is signal-independent noise. The PRNU is due to the
difference of response of each pixel to a given signal; it is
Figure 3. MTF experimental values of the scientific camera CCD detector
therefore signal-dependent noise.
at different wavelengths of the visible spectrum.
The FPN was corrected by subtracting from the speckle
image the dark image captured obscuring the detector, and the
PRNU by means of the procedure proposed elsewhere [15].
For the processing of the speckle images, the appropriate
software was developed using MATLAB.

Results and Discussion

For each wavelength analysed, the experimental values of
the horizontal MTF of the detectors were calculated using Eq.
(1).
For the three detectors, at each wavelength analysed, a
polynomial fit of the experimental MTF values was made and
the resulting functional expression was normalized by dividing it
by the value that the adjustment equation provided at zero
frequency.
The MTF experimental values of the detectors were
normalized by dividing them by the same value used to
normalize the corresponding adjustment curve (zero-order
coefficient in the polynomial-fit expression). Figure 4. MTF experimental values of the scientific camera CMOS
The results are shown in Figures 2, 3, and 4, which reflect, detector at different wavelengths of the visible spectrum.
for the different wavelengths of the visible spectrum, the

CGIV 2008 and MCS’08 Final Program and Proceedings 463

 In the case of the CCD video camera, considering that by
definition the MTF is normalized at unity at zero spatial
frequency, MTF values higher than one are reached due to the
amplification introduced by the electronic filters of the image
card [24, 25].
Figures 5, 6, and 7 show, for the three detectors and at the
different wavelengths, the MTF curves given by the polynomial
adjustments of the experimental values, after normalization at
zero spatial frequency.
For each wavelength, the MTF curve of the low-cost video
camera CCD detector was determined by fitting experimental
values to a third-order polynomial function. The different
correlation coefficients associated with these fits are listed in
Table 4.

Figure 6. MTF of the scientific camera CCD detector at different

wavelengths of the visible spectrum. Curves were determined by fitting
experimental values to a second-order polynomial function.

Table 5: Correlation coefficients of fitting curves shown in

Figure 6
Wavelength (nm) Correlation coefficient

514 0.9646

488 0.9717

Figure 5. MTF of the low-cost video camera CCD detector at different 457 0.9727
wavelengths of the visible spectrum. Curves were determined by fitting
experimental values to a third-order polynomial function.
For the scientific camera CMOS detector, MTF curves
were determined by fitting experimental values to a third-order
polynomial function at each wavelength. The correlations
Table 4: Correlation coefficients of fitting curves shown in coefficients associated with these polynomial fits are shown in
Figure 5 Table 6.
Wavelength (nm) Correlation coefficient

514 0.9847

502 0.9929

488 0.9947

477 0.9954

454 0.9958

At the different wavelengths, MTF curves of the scientific

camera CCD detector were determined by fitting experimental
values to a second-order polynomial function. The correlation Figure 7. MTF of the scientific camera CMOS detector at different
coefficient corresponding to each of these adjustments is wavelengths of the visible spectrum. Curves were determined by fitting
indicated in Table 5. experimental values to a third-order polynomial function.

464 ©2008 Society for Imaging Science and Technology

Table 6: Correlation coefficients of fitting curves shown in studied. Also, the differences between the MTF curves
Figure 7 corresponding to the different wavelengths become more notable
Wavelength (nm) Correlation coefficient as the spatial frequency increases.
In the case of scientific cameras, except at spatial
514 0.9904 frequencies near the Nyquist frequency, no significant
differences are appreciated in the MTF of the detector at the
488 0.9913 different wavelengths, within the range studied.
Moreover, results prove the scientific CCD detector
presented MTF values higher than those of the CMOS detector
457 0.9819
at the same spatial frequencies, for each of the three visible
wavelengths analysed.
With the CCD video camera, the MTF reached values
The comparison of the results show the differences between higher than one due to the amplification introduced by the
the cameras analysed with respect to the performance of the electronic filters of the image card.
MTF of the detector with wavelength, within the spectral range Since the penetration depth of photons in the detector
studied. material increases with the wavelength, effects of smear and
For the CCD detector of the video camera, the highest signal loss are more pronounced at higher wavelengths.
value of the MTF was reached at the lowest of the wavelengths Therefore, the influence of the diffusion on the MTF of the
studied. The value of the spatial frequency for which the detector is stronger as the wavelength increases. In this sense, it
maximum of the MTF curve is found diminishes as the would be worthwhile to carry out new measurements at higher
wavelength augments. Furthermore, the differences between the wavelengths than those analysed here.
MTF curves corresponding to the different wavelengths At the moment, we are interested in extending the present
analysed become more notable as the spatial frequency work to study the MTF spectral-variation of different detector
increases. arrays at higher wavelengths within the visible spectrum.
In the case of scientific cameras, the behaviour of the MTF
with wavelength does not present the same trend as that Acknowledgements
observed for the low-cost video camera. Except at spatial The authors express their appreciation to the Ministerio de
frequencies close to the Nyquist frequency, for the CCD and Ciencia y Tecnología of Spain for financing project FIS2004-
CMOS detectors, no significant differences are appreciated in 06465-C02-02
the MTF curves resulting at the different wavelengths, within
the range of the visible spectrum studied. References
For each of the three wavelengths studied, the CCD [1] G. C. Holst and T. S. Lomheim, CMOS/CCD Sensors and Camera
detector presented MTF values higher than those of the CMOS Systems (JCD Publishing, Winter Park, FL. and SPIE Optical
detector. Engineering Press, Bellingham, WA, 2007).
Differences in the MTF spectral behaviour of the two CCD [2] H. Helmers and M. Schellenberg, “CMOS vs. CCD sensors in
detectors could be due to the effect of charge diffusion between speckle interferometry”, Opt. Laser Technol., 35, 587–595 (2003).
pixels, which depends on wavelength [24]. Probably, the charge [3] D. Litwiller, “CCD vs. CMOS: facts and fictions”, Photonics
diffusion effect is greater for the low-cost CCD video camera, Spectra, 154–158 (2001).
and therefore wavelength influences the MTF more significantly. [4] J. Janesick, “Dueling detectors. CMOS or CCD?”, SPIE’s OE
In the case of the video camera, the horizontal MTF values are Magazine, 41, 30-33 (2002).
also affected by the electronic filters of the image card [11, 25]. [5] J. Janesick, “Lux transfer: complementary metal oxide
For both scientific detectors, the overall MTF behaviour is semiconductors versus charge-coupled devices”, Opt. Eng., 41,
determined by the pixel active area geometrical shape and the 1203–1215 (2002).
physical diffusion effect [24, 26]. The diffusion component of the [6] G. Deptuch, A. Besson, P. Rehak, M. Szelezniak, J. Wall, M.
MTF is due to the penetration depth of photons in the substrates Winter and Y. Zhu, “Direct electron imaging in electron
and, as the wavelength increases, photon absorption occurs at microscopy with monolithic active pixel sensors”,
increasing depths in the detector material [24, 26]. It is therefore Ultramicroscopy, 107, 674–684 (2007).
expected that the MTF of the scientific detectors changes more [7] S. K. Park, R. Schowengerdt and M. Kaczynski, “Modulation-
clearly at longer wavelengths than those analysed in the present transfer-function analysis for sampled image system”, Appl. Opt.,
work. 23, 2572-2582 (1984).
[8] J. C. Feltz and M. A. Karim, “Modulation transfer function of
Conclusions and Future Work charge-coupled devices”, Appl. Opt., 29, 717-722 (1990).
In this work, we have comparatively analysed the quality of [9] D. N. Sitter, Jr., J. S. Goddard, and R. K. Ferrell, “Method for the
the images provided by different detector arrays using the measurement of the modulation transfer function of sampled
speckle method. In addition, we have compared their imaging systems from bar-target patterns”, Appl. Opt., 34, 746-751
performance with wavelength within a range of the visible (1995).
spectrum. For this, we have studied the MTF at several visible [10] A. Daniels, G. D. Boreman, A. D. Ducharme, and E. Sapir,
wavelengths, for the detectors of a low-cost CCD video camera “Random transparency targets for modulation transfer function
and of two scientific cameras (a CCD and a CMOS). measurement in the visible and infrared regions”, Opt. Eng., 34,
Our results reveal differences in the MTF spectral-variation 860-868 (1995).
of the detector arrays analysed within the spectral range studied. [11] S. M. Backman, A. J. Makynen, T. T. Kolehmainen, and K. M.
For the CCD detector of the video camera, the highest Ojala, “Random target method for fast MTF inspection”, Opt.
value of the MTF was reached at the lowest of the wavelengths Express, 12, 2610-2615 (2004).

CGIV 2008 and MCS’08 Final Program and Proceedings 465

[12] N. Guérineau, J. Primot, M. Tauvy and M. Caes, “Modulation methods to evaluate image quality of systems based on CCD and CMOS
transfer function measurement of an infrared focal plane array by detectors, publishing several papers in these fields.
use of the self-imaging property of a canted periodic target”, Appl.
Opt., 38, 631-637 (1999). Manuel Rubiño received his M.S. (1986) and Ph.D. in Physics
(1990) from the University of Granada, Granada, Spain. Since then he
[13] M. Marchywka and D. G. Socker, “Modulation transfer function
has worked as an associate professor of Radiometry, Photometry,
measurement techniques for small-pixel detectors”, Appl. Opt., 31,
Colorimetry and Optical Technology in the Department of Optics at the
7198-7213 (1992).
University of Granada. His research interests are optical and
[14] J. E. Greivenkamp and A. E. Lowman, “Modulation transfer radiometric characterization of CCD and CMOS cameras, applied
function measurements of sparse-array sensors using a self- Colorimetry and applied Photometry.
calibrating fringe pattern”, Appl. Opt., 33, 5029-5036 (1994).
[15] A. M. Pozo, A. Ferrero, M. Rubiño, J. Campos and A. Pons,
“Improvements for determining the modulation transfer function of
charge-coupled devices by the speckle method”, Opt. Express, 14,
5928-5936 (2006).
[16] G. D. Boreman and E. L. Dereniak, “Method for measuring
modulation transfer function of charge-coupled devices using laser
speckle”, Opt. Eng., 25, 148-150 (1986).
[17] A. D. Ducharme, "Microlens diffusers for efficient laser speckle
generation", Opt. Express, 15, 14573-14579 (2007).
[18] G. D. Boreman, Y. Sun and A. B. James, “Generation of laser
speckle with an integrating sphere”, Opt. Eng., 29, 339-342 (1990).
[19] M. Sensiper, G. D. Boreman, A. D. Ducharme and D. R. Snyder,
“Modulation transfer function testing of detector arrays using
narrow-band laser speckle”, Opt. Eng., 32, 395-400 (1993).
[20] A. M. Pozo and M. Rubiño, “Comparative analysis of techniques
for measuring the modulation transfer functions of charge-coupled
devices based on the generation of laser speckle”, Appl. Opt., 44,
1543-1547 (2005).
[21] J. W. Goodman, Statistical properties of laser speckle and related
phenomena, in Laser Spekle and Related Phenomena, (J. C. Dainty,
Ed., Vol. 9 of Topics in Applied Physics Springer-Verlag, Berlin,
1984), pg. 35-40.
[22] L. I. Goldfischer, “Autocorrelation function and power spectral
density of laser-produced speckle patterns”, J. Opt. Soc. Am., 55,
247-253 (1965).
[23] G. D. Boreman, “Fourier spectrum techniques for characterization
of spatial noise in imaging arrays”, Opt. Eng., 26, 985-991 (1987).
[24] G. C. Holst, CCD Arrays, Cameras and Displays, (JCD Publishing,
Winter Park, FL. and SPIE Optical Engineering Press, Bellingham,
WA, 1996).
[25] B. T. Teipen and D. L. MacFarlane, “Liquid-crystal-display
projector-based modulation transfer function measurements of
charge-coupled-device video camera systems”, Appl. Opt., 39, 515-
525 (2000).
[26] I. Shcherback and O. Yadid-Pecht, CMOS APS MTF Modeling, in
CMOS Imagers: From Phototransduction to Image Processing, (O.
Yadid-Pecht and R. Etienne Cummings, Eds., Kluwer Academic,
Norwell, MA, 2004), pg. 53-74.

Author Biography
Alicia Fernández-Oliveras received the B.S. degree in physics from
University of Granada, Granada, Spain, in 2007 and she is pursuing the
M.S. degree in physics. She currently collaborates with the Department
of Optics, University of Granada, where she has developed several
research contracts under project FIS2004-06465-C02-02, financed by
the Spanish Ministry of Education. She is engaged in the research,
development, and applications of measurement methods to evaluate
image quality of CCD and CMOS arrays systems.

Antonio M. Pozo received the B.S. and M.S. degrees in physics from
University of Granada, Granada, Spain, in 2000 and 2003, respectively,
where he is also a graduate in optics and optometry. He is an associate
professor with the Department of Optics in the Science Faculty,
University of Granada. He has been conducting research in experimental

466 ©2008 Society for Imaging Science and Technology