Optics and Photonics Journal, 2013, 3, 250258 http://dx.doi.org/10.4236/opj.2013.33040 Published Online July 2013 (http://www.scirp.org/journal/opj) Recognition of Direction of New Apertures from the Elongated Speckle Images: Simulation Abdallah Mohamed Hamed Physics Department, Faculty of Science, Ain Shams University, Cairo, Egypt Email: Amhamed73@hotmail.com Received January 31, 2013; revised March 2, 2013; accepted March 9, 2013 Copyright © 2013 Abdallah Mohamed Hamed. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. ABSTRACT In this paper, we present an elongated speckle images produced from diffusers using sharp elliptical apertures. The ori entation of the elliptic aperture is recognized from the direction of the elongation in the speckle images. The aperture tilting out of the plane is investigated. Three models of elliptical apertures are considered and the corresponding speckle images are obtained. The 1st model is composed of two orthogonal ellipses or plus symbol pupil; the 2nd has four sym metric ellipses with an angle of 45˚ between each of them or in the form of a snow flake pupil and the 3rd model looks like an airplane. Also, the autocorrelation profiles of the speckle images corresponding to the diffused airplane are ob tained from which the average speckle size is computed. Finally, the reconstructed images of the described elliptical models and its autocorrelation images, making use of Mat lab code, are obtained. Keywords: Digital Imaging; Sharp Elliptical Apertures; Plus Symbol and Snow Flake Pupils 1. Introduction The production of elongated speckle images was ob tained using the mechanical scanning of the static spec kle pattern [1]. The author presented a technique of spa tially oriented specklescreen encoding to improve the grating encoding technique for whitelight image proc essing. Also, an artificial screen composed of small strips photographed several times on a high resolution film is designed to obtain elongation of nearly ten times the av erage grain size of natural speckle [2]. In a recent publi cation by the author [3], numerical elliptical apertures of small elliptic shapes are analyzed and the Fourier trans form is operated giving the speckle images of diffusers modulated by these elliptic apertures. An approach for determining the roughness of engi neering surfaces is resulted from the speckle elongation effect. The laser speckle pattern, arising from light scat tered from rough surfaces that are illuminated by poly chromatic laser light, is detected in the farfield region. The incoherent superposition of these light intensities and the angular dispersion cause the effect of speckle elongation [4,5]. This is characterized by increasing speckle widths and leads to a radial structure of the spec kle patterns. With increasing surface roughness, the elon gation is replaced more and more by the decorrelation of the monochromatic speckle patterns for the different wavelengths. Such effects were detected with the CCD technique and analyzed by local autocorrelation func tions of intensity fluctuations that were calculated for different areas of the speckle patterns. Hence, the auto correlation method is applied to process laser speckle patterns. The relation between surface roughness, speckle elongation, and correlation length of autocorrelation function can be obtained. Consequently, the measured surface roughness can be achieved [6]. An oriented pho tographic diffuser was used to record an elongated spec kle pattern. It is found that the contrast transfer, when gratings are imaged through the slits in the diffuser, is considerably higher compared to imaging through a cir cular pinhole of comparable dimensions [7]. An autocor relation algorithm for speckle size evaluation has been investigated [811]. The authors measured the average speckle size from the autocovariance function of the digitized intensity speckle pattern. The spatial character istics such as speckle size can be used to measure the roughness of surfaces [12,13]. An important remark must be taken into consideration during the recording of speckle data which states that the speckle size must be greater than the pixel size of the CCD camera in order to resolve variations in speckle intensity [14]. Recently, a Fast cal culation method for optical diffraction on tilted plane was investigated [15]. C opyright © 2013 SciRes. OPJ
A. M. HAMED 251 In this work, two sharp elliptical apertures with dif ferent orientations are proposed and the corresponding elongated speckle images are obtained. Also, three dif ferent new models in the form of the plus symbol, a snowflake and an airplane, are numerically simulated and the corresponding speckle images are computed and plotted. Finally, the reconstructed images of the models and the autocorrelation of these models are plotted. It is noted that the sharp ellipses considered in this study have a semimajor axis ten times larger than the semiminor axis as compared with the recent work [3] with the semi major axis nearly equal to the semiminor axis. These new models allow the recognition of the different elliptic models from the speckle images as compared with the circular aperture. 2. Formation of Speckle Images Using Diffusers Modulated by Different Sharp Elliptic Apertures In this work, a simple ellipse of semimajor axis ten times the semiminor axis is investigated. We assume that this sharp ellipse is represented as follows: 22 22 ,1; 1with0.1 xy xyb a ab (1) The numerical ellipse is made from pixels of dimen sions 1024 1024 and is represented as follows: ,1; ,1024,1024fmxny mn (2) ∆x and ∆y are unity. A diffuser of random phase changes, has the same di mensions like the above described elliptical aperture, is written as follows: 2 ,exp and,; k dxyjr xyk f (3) In matrix form, the simulated diffuser becomes: ,exp and, k d m xn yjrm xn y f , (3′) Consider coherent imaging system, with laser uniform illumination incident upon the object (the diffuser) fol lowed by the simulated sharp elliptic aperture. In this case, we can write the complex amplitude transmittance as follows: ,,cmxnyf mxnydmxny (4) Hence, the Fourier transform of Equation (4), realized by a converging lens, gives in its front focal plane this complex amplitude: , mn mn cf d where .. , FTfmx ny ..d ,T mxny , dF , and . .,cFTcmxny The symbol stands for the convolution product. Equation (5) means that the modulated elongated speckles are formed from the convolution product of both of the Fourier spectrum of the diffuser and the sharp elongated aperture. The direction of speckle is elongated normal to the semimajor axis of the elliptic aperture. 2.1. Effect of the Aperture Tilting upon the Elongated Speckles Assume that the aperture is tilted making an angle with the aperture plane (x, y) which is not a simple trans lation. In this case, the sharp elliptic aperture is decom posed into two components one along the optic axis z and the other lies in the plane (x, y) assumed along ydirec tion written as follows: exp cos, tilted jkyfx y (6) The concept of aperture tilting which is represented as a linear phase shift in the Fourier plane resembles the inclined plane wave which gives a linear shift in its front focal plane. Hence, the coherent light emitted from laser beam is incident upon the tilted aperture described by Equation (6) and the transmitted light is incident upon the diffuser described in the precedent section Equation (3) giving this complex amplitude: ,, tilted tilted cmxnyfmxny dmxny, (7) Substitute from Equation (6) in Equation (7), we re write Equation (7) as follows: , exp cos,, tilted cmxny jkmyf mxnydmxny (8) Applying the Fourier transform operation and making use of convolution operation we get: .. cos ,, , ctilt apert mn mnmnf fd xyxyx y (9) represents the DiracDelta distribution. This Delta function displaces the whole speckle pat tern by an amount equal to cosf in the direction conjugate to the y direction since f is the focal length of the Fourier transform lens. Hence, for the tilted aperture the complex amplitude of the speckle pattern is written as follows: cos .. , mn f ctilt apertcxy (10) , yx y (5) Copyright © 2013 SciRes. OPJ
A. M. HAMED 252 It is noted that the linear shift of the whole speckle images is large as compared with the field of view of the whole image. To detect the shift this condition must be fulfilled: fcos( < 3 ; is the average speckle size. 2.2. Autocorrelation Algorithm for Speckle Size Evaluation The average speckle size of a speckle image is estimated by calculating the autocovariance function of the digi tized intensity speckle pattern as follows: If I(x1, y1) and I(x2, y2) represent the intensities of two points in the imaging plane (x, y), the intensity autocor relation function is defined by equation 112 2 ,,, I cxy IxyIxy (11) where 12 xx , 12 yy y and corresponds to a spatial average. The autocovariance function corresponds to the nor malized autocorrelation function of the intensity which has zero base and its full width at half maximum (FWHM) provides a reasonable measurement of the average width of a speckle [6]. In order to use autocorrelation function method to calculate the average speckle size, it requires sufficient sampling speckles in an image to give a rea sonable statistical evaluation. A Matlab program is used [8] to compute the auto covariance of the speckle image. The calculated auto covariance functions are shown and the FWHM of the calculated function gives the average speckle size of the speckle pattern. 2.3. The Reconstruction Process and the Autocorrelation of Elliptic Apertures Since the complex amplitude of modulated speckle im ages using the above different apertures is given by Equ ation (5) as follows: ,, , ,, mn mn cuv fd yx fuv duv y (5′) ,, mn y uv are the reduced coordinates in the Fourier plane of speckles. The reconstruction of the different apertures obtained by operating the inverse Fourier transform upon Equation (5) to get: 1 ,.., ,Rx yFTfuvduv (12) , y are the Cartesian coordinates in the imaging reconstruction plane. Making use of the properties of Fourier transform and convolution, then we get finally this result: ,, Rx yfx ydx y , (13) Hence, in the imaging plane we localize the aperture image affected by a noise originated from the diffuser function. The autocorrelation function of the different apertures is obtained by operating the Fourier transform upon the intensity of the speckle image as follows. Firstly we get the speckle intensity as the modulus square of the complex amplitude of the speckle Equation (5') as follows: 2 ,, uvc uv (14) Operating the F.T‒1 over Equation (14), we get the au tocorrelation function of the multiplication product as follows: 1 2 1 ,.., .. ,, , ,,, cx yFTIuv FTfu vdu v xydxyfxydxy (15 3. Results and Discussion nges used in the forma diffuser is multiplied by the sharp elliptical aper tu ) A diffuser of randomly phase cha tion of all modulated speckle images is shown in Figure 1. The re of semimajor axis ten times the semiminor axis as shown in the left of Figure 2. A matrix of dimensions 1024 × 1024 pixels for this diffused aperture is consid ered. The elongated speckle pattern is obtained by oper ating the FFT algorithm upon the diffused aperture as 100 200 300400 500600 700 800900 1000 100 200 300 400 500 600 700 800 900 1000 Figure 1. A random diffuser d(x,y) = exp[(j 2/f) rand(x with dimensions 1024 × 1024 pixels. ,y)] Copyright © 2013 SciRes. OPJ
A. M. HAMED 253 Figure 2. On the left, an elliptical aperture of a semimajor axis a = 2 and semiminor axis b = 0.2 with matrix dimen shown that the longation of speckles is normal to the major axis of the angle = 45˚ with the xaxis and the co α is the angle between the inci de sions of 1024 × 1024 pixels is considered. On the right is the corresponding speckle image elongated along the ydirec tion of matrix dimensions 256 × 256 pixels. The elliptical aperture is superimposed over the diffuser and its semi major axis is extended along xdirection. shown on the right of Figure 2. It is e ellipse shown along xdirection giving the elongation directed along the yaxis. In my opinion, this can be in terpreted if we consider that the elongated elliptical ap erture looks like a slit of finite width along ydirection and hence its Fraunhoffer diffraction pattern located in the Fourier plane is represented by approximate sinc(y) function varies along y. Consequently, the convolution product of the speckle pattern formed from the ordinary diffuser and the diffraction pattern of the elongated el lipse will give the elongated speckle distribution shown along ydirection as in Figure 2. This reasoning is estab lished from the consideration of sharp elliptic aperture of minor axis equal to the width of the slit along ydirection while the major axis equal to the length of the slit along xdirection. Based, on this analogy all the proposed mo dels are justified. Figure 3 shows the diffused sharp ellipse with its ma jor axis making an rresponding elongated speckle with the elongation orthogonal to the pupil major axis. It is shown that the direction of the ellipse is recognized from the direction of the elongated speckle where both of them must be or thogonal to each other. The effect of aperture misalignment is shown in Fig ure 4. The tilting angle nt ray and the normal to the aperture plane. This angle is taken to be 3˚ and 4˚ as shown in figures (b), (c) and compared with the case of perfectly aligned aperture shown in Figure 4(a). This tilting is completely different as compared with aperture orientation in its plane (results Figure 3. On the left, an elliptical aperture making an angle 0f 45˚ with the xaxis and of matrix dimensions of 1024 × 1024 pixels is shown. On the right is the corresponding speckle image elongated normal to the semimajor axis of matrix dimensions 256 × 256 pixels. The elliptical aperture is superimposed over the diffuser. 50100150200250 50 100 150 200 250 50100 150 200 250 50 100 150 200 250 (a) (b) 50 100 150 200 250 50 100 150 200 250 (c) Figure 4. Three different elongated speckle imes using the sharp elliptic aperture show Figure 2. (a) On the left is e three elon ated speckle images taking the aligned aperture as a ag n in the elongated speckle for aligned sharp elliptic aperture. While (b) and (c) are obtained for misaligned aperture tak ing the tilting angle α = 3˚ and 4˚ respectively. shown in Figures 1 and 2). If we compare th g reference hence we can predict any misalignment due to the aperture tilting by examining the speckle pattern. Any larger inclinations will give different speckle patterns since the inclination in the pupil plane is transformed into Copyright © 2013 SciRes. OPJ
A. M. HAMED 254 linear shift in the Fourier plane. Hence, larger aperture tilting give a difference as compared with the aligned aperture but small tilting is interesting to check the alignment of the optical system. Also, larger aperture tilting give a difference as com pared with the aligned aperture but small tilting is inter es ft of Fi dimensions 2048 × orresponding to ting to check the alignment of the optical system. The plus symbol pupil is composed of two orthogonal ellipses modulated by the diffuser is shown in the le gure 5 for a matrix of 2048 × 2048 pixels. On the right is the corresponding speckle image. The elongation of the speckle image is shown directed along both of the x and yaxes. Also, another arrangement of a snow flake or four ellipses superimposed over the diffuser and the cor responding elongated speckle image is plotted as in Fig ure 6. This arrangement shows a complicated elonga tion since each ellipse gives speckle elongation orthogo nal to its elliptical major axis. Hence, the recognition of the snow flake pupil is attributed to the four different directions shown for the elongation of the same speckle. All the elongated speckle images shown in Figures 2 6 are of dimensions 256 × 256 pixels. The 3rd model of a pupil in the form of an airplane su perimposed over the diffuser of matrix 2048 pixels is numerically constructed as shown in Figure 7. The corresponding speckle image is obtained by operating the FFT showed a specific elongation along the plane (x, y) represented as in Figure 8. The profile shape of the autocorrelation intensity is shown as in Figure 9(a), along the xaxis c the elongated speckles Figure (8) for the airplane pu pilof 2048 2048 pixels. Also, the autocorrelation pro file along the yaxisis shown in Figure 9(b). The spec Figure 5. On the left, a snow flake pupil of two orthogona ellipses of matrix dimensions 2048 × 2048 pixels is shown l . On the right is the corresponding speckle image elongated along the x and ydirections and the matrix dimensions are 256 × 256 pixels. The elliptical aperture is superimposed over the diffuser. Figure 6. On the left, a snow flake pupil superimposed ove the diffuser of matrix dimension 2048 × 2048 pixels i r s shown. On the right is the corresponding speckle image elongated along the different directions of matrix dimen sions 256 × 256 pixels. Figure 7. A pupil in the form of an airplane superimpose over the diffuser of matrix dimensions 2048 × 2048 pixels. d Figure 8. Speckle elongation in different directions of di mensions 256 × 256 pixels corresponding to the airplane diffused pupil of dimensions 2048 × 2048 pixels. Copyright © 2013 SciRes. OPJ
A. M. HAMED 255 (a) (b) Figure 9. (a) Autocorrelationtensity along the x – axis of the elongated speckles shown Figure (8) for the airplane is computed by taking the full width at half in in pupil of 2048 × 2048 pixels. The average speckle size is x = (4.5 mm/512 pixels) (5 pixels) = 44 mm; (b) Autocorrelation intensity along the yaxis of the elongated speckles shown in Figure (8) for the airplane pupil of 2048 × 2048 pixels. The average speckle size is y = (3.6 mm/512 pixels) (8 pixels) = 56 mm. kle size maximum along xdirection, referring to Figure 9(a) as follows: FWHM = ∆x = 5 pixels and along the ydirec tion, referring to Figure 9(b) as: ∆y = 8 pixels. Hence the average speckle size along xdirection is calculated, as suming field of view in the speckle image equal 4.5 mm 3.6 mm, as follows: 4.5 mm512 pixels5 pixels44mm x while the average speckle size along the ydirection is computed as: 3.6 mm512 pixels8 pixels56 mm y For comparison, the average speckle size for circular apert Fi ure of radius 128 pixels is obtained (referring to gures 1012 as follows): Figure 10. uniform circular aperture of pixels dimensions 2048 × 2048 and radius 128 pixels. Figure 11. The field of view is 4.5 mm × 3.6 mm correspond ing to 512 × 512 pixels for the speckle image, and is ob tained for the diffuser provided with uniform circular ap erture shown in Figure 10. ixels10 pixels44m 4.5 mm2048 p x 3.6 mm2048 pixels10 pixels35m y Another set of three different speckle images using a circular aperture are plotted as in Figure (13). On th is e left the speckle for the aligned circular aperture. While the other two figures (b), (c) are obtained for misaligned cir cular apertures, taking the tilting angles α =3˚ and 4˚ re spectively. All apertures have equal radii of 64 pixels. The autocorrelation intensity of the corresponding spec kle patterns are shown as in Figures 14(a)(c). It is shown that the aligned aperture is recognized from its speckle pattern Figure 14(a) since it is different from the tilted aperture Figures 14(b) and (c). Hence, any misalignment due to aperture tilting is recognized referring to its dif ferent speckle pattern. Copyright © 2013 SciRes. OPJ
A. M. HAMED Copyright © 2013 SciRes. OPJ 256 (a) (b) Figure 12. (a) Autocorrelation intensity of the speckle imageown in Figure 10 along the xaxis. The FWM of the speckle size corresponding to the 10 pixels sh4.5 mm/1024 pixels)(10 pixels) = 44 shH own on the autocorrelation peak is equal to: s = ( mm;(b) Autocorrelation intensity of the speckle image shown in Figure 9 along yaxis. The FWHM of the speckle size corre sponding to the 10 pixels shown on the autocorrelation peak is equal to: s = (3.6 mm/1024 pixels)(10 pixels) = 35 mm. 20 40 60 80100120 20 40 60 80 100 120 20 406080100120 20 40 60 80 100 120 20 406080100120 20 40 60 80 100 120 (a) (b) (c) Figure 13. Three different speckle images using a circular aperture. On the left is the specklee aligned circular aperture. While the other two figures (b), (c) ae tilting angles α = 3˚ and 4˚ re for th re obtained for misaligned circular apertures, taking th spectively. All apertures have equal radii of 64 pixels. 050 100 150 200250 300 0.01 0 0. 01 0. 02 0. 03 0. 04 0. 05 0. 06 0. 07 050 100150 200 250300 0.0 1 0 0. 01 0. 02 0. 03 0. 04 0. 05 0. 06 0. 07 pixel s al ong xaxi s autocorrel atio n i nt ensit y for al i gned ci rcul ar apert pix els al ong xax i s aut ocorrel a t i on i nten s i ty for m i sal i g ned aper t ure al ph a= 3 (a) (b) ure 050100 150 200250 300 0.01 0 0.01 0.02 0.03 0.04 0.05 0.07 0.06 pix el s al ong xaxis autocorrelation int ensi ty for m i saligned apert ure (c) Figure 14. (a) The autocorrelation intensity of the speckle pattern corresponding to the aligned circular aperture. (b) The autocorrelation intensity of the speckle pattern correspondin misaligned circular aperture of tilting angle = 3˚. (c) The autocorrelation intensity of the speckle pattern corresponding to misaligned circular aperture of tilting angle = 4˚. alpha= 4 g to
A. M. HAMED 257 The reconstruction of the two crossed ellipses using the inverse Fourier transform of matrix dimensions 2048 2048 pixels is shown as in Figure 15. Also, the recon × st nd the autocorrelation of the four eq oposed sharp models is for the on by comparing the different speckle inary speckle image for uniform cir ructed image of the aperture of four crossed ellipses is shown as in Figure 16. The autocorrelation of the two orthogonal ellipses ob tained from the reconstruction of the speckle intensity is shown as in Figure 17 a ually spaced ellipses where the angle between each two is = 45˚ are shown as in Figure 18. The autocorre lation images shown in Figures 17 and 18 are obtained using Equation (15). 4. Conclusions The motivation of the pr sake of its recogniti images with the ord cular aperture and to check its alignment in its plane. The snowflake and the airplane models showed a rela tively complicate elongation since each part give elonga Figure 15. Reconstruction of the plus symbol pupil of ma trix 2048 × 2048 pixels. Figure 17. The autocorrelation of plus symbol pupil or the two orthogonal ellipses. Figure 18. The autocorrelation of the snow flake pupil or the four crossed ellipses. The alignment of optical microscopic systems can be verified by testing the aperture inclination using speckle techniques. A potential application of this work may be extended to polychromatic illumination using a mixture of HeNe laser and Ar ion laser in order to differentiate the colored parts of the elliptic apertures. REFERENCES [1] G. G. Mu, Z. Q. Wang, Q. Gong, Q. W. Song and F. X. Wu, “WhiteLight Image Processing Using Oriented Spec kleScreen Encoding,” Optics Letters, Vol. 10, No. 8, 1985, pp. 375377. doi:10.1364/OL.10.000375 tion normal to the concerned elliptical part allowing ain its recognition. ag lied to an Optical Processor Using Speckle Techniques,” Journal of Modern Optics , Vol. 38, [2] N. Barakat, A. M. Hamed , H. El Ghandoor, K. ElDo himy, M. A. Fadly and O. A. Abdel Ghafar, “A Photo graphic Encoder App Figure 16. Reconstruction of the snow flake pupil of matrix 2048 × 2048 pixels. No. 1, 1991, pp. 203208. Copyright © 2013 SciRes. OPJ
A. M. HAMED 258 doi:10.1080/09500349114550221 [3] 1, 2011, pp. 17. [4] A. M. Hamed, “Discrimination between Speckle Images Using Deformed Apertures,” Optical Engineering, Vol. 56, No. P. Lehmann, S. Patzelt and A. Schöne, “Surface Rough ness Measurement by Means of Polychromatic Speckle Elongation,” Applied Optics, Vol. 36, No. 10, 1997, pp. 21882197. doi:10.1364/AO.36.002188 [5] P. Lehmann, “Aspect Ratio of Elongated Polychromatic FarField Speck Distribution woughness Charac les of Continuous and Discrete Spectral ith Respect to Surface R terization,” Applied Optics, Vol. 41, No. 10, 2002, pp. 2008 2014. doi:10.1364/AO.41.002008 [6] Z. H. Yuan, et al., “Comparison of Surface Roughness Measurement in Dichromatic Speckle Patterns with Auto correlation Method,” Advanced Materials Research, Vol. 189, 2011, pp. 680683. [7] A. S. Kumar and R. M. Vasu, “Imaging with Oriented Photographic Diffusers,” Instrumentation and Services Unit, Indian Institute of Science, Bangalore, 2002. [8] H. Lin, “Speckle Mechanism in Holographic Optical Co kle Mechanism in Holographic herence Imaging,” Ph.D. Dissertation, University of Mis souri, Kansas City, 2009, pp. 5758. [9] H. Lin and P. Yu, “Spec Optical Imaging,” Optics Express, Vol. 15, No. 25, 2007, pp. 1632216327. doi:10.1364/OE.15.016322 [10] J. W. Goodmann, “Statistical Properties of Laser Speckle Patterns in Laser Speckle and Related Phenomena,” Sprin gerVerlag, New York, 1984. [11] Y. Piederriere, J. Le Meur, J. Cariou, J. Abgrall and M. 19, 2004, pp. 4596 Blouch, “Particle Aggregation Monitoring by Speckle Size Measurement, Application to Blood Platelets Aggrega tion,” Optics Express, Vol. 12, No. 4601. doi:10.1364/OPEX.12.004596 [12] P. Lehmann, “Surface Roughness Measurement Based on the Intensity Correlation Function of Scattered Light un der Speckle Pattern Illumination,” Applied Optics, Vol. 38, No. 7, 1999, pp. 11441152. doi:10.1364/AO.38.001144 [13] R. Berlasso, F. Perez Quintian, A. M. Rebollo, A. C. Raffo and G. N. Gaggioli, “Study of Speckle Size of Light Scattered from Cylindrical Rough Surfaces,” Applied Op tics, Vol. 39, No. 31 , 2000, pp.58115819. doi:10.1364/AO.39.005811 [14] L. T. Alexander, E. J. Harvey and R. A. Weeks, “Average Speckle Size as a Function of Intensity Threshold Level: Comparison of Experimental Measurements with The ory,” Applied Optics, Vol. 33, No. 35, 1994, pp. 8240 8250. doi:10.1364/AO.33.008240 [15] K. Matsushima, H. Schimmel and F. Wyrowski, “Fast Calculation Method for Optical Diffraction on Tilted Planes by Use of Angular Spectrum of Plane Waves,” Journal of the Optical Society of America, Vol. 20, No. 9, 2003, pp. 17551762. doi:10.1364/JOSAA.20.001755 Copyright © 2013 SciRes. OPJ
