Twin image elimination in digital holography by combination of Fourier transformations
AAccepted in Journal of Optics (2013) of Optical Society of India (authors’ accepted version)
Twin image elimination in digital holography bycombination of Fourier transformations
Debesh Choudhury · Gautam LoharAbstract
We present a new technique for removing twin image in in-line digitalFourier holography using a combination of Fourier transformations. Instead ofrecording only a Fourier transform hologram of the object, we propose to record acombined Fourier transform hologram by simultaneously recording the hologram ofthe Fourier transform and the inverse Fourier transform of the object with suitableweighting coefficients. Twin image is eliminated by appropriate inverse combinedFourier transformation and proper choice of the weighting coefficients. An opticalconfiguration is presented for recording combined Fourier transform holograms.Simulations demonstrate the feasibility of twin image elimination. The hologramreconstruction is sensitive to phase aberrations of the object, thereby opening away for holographic phase sensing.
Keywords
Holographic twin image, in-line digital holography, combined Fouriertransform.
Twin image is an age old problem in holography since its invention by Gabor [1].In 1951, Bragg and Rogers first eliminated the unwanted twin image by record-ing two holograms and doubling the object-to-hologram distance in between theexposures [2]. A nice review details several available techniques for getting rid ofthe twin image [3]. The off-axis holography of Leith and Upatnieks is the sim-plest method, but it requires high resolution recording materials [4]. Most othermethods [3] either used optical/digital spatial filtering, or needed to record multi-ple phase shifted holograms, or utilized iterative reconstruction of the holograms
Debesh ChoudhuryDepartment of Electronics and Communication EngineeringNeotia Institute of Technology, Management and SciencePO - Amira, D. H. Road, South 24 Parganas, Pin 743368, West Bengal, IndiaE-mail: debesh[AT]iitbombay[DOT]orgGautam LoharDepartment of Electronics and Communication Engineering, JIS College of EngineeringBlock A, Phase III, Kalyani, Nadia, Pin 741235, West Bengal, India. a r X i v : . [ phy s i c s . op ti c s ] N ov Debesh Choudhury, Gautam Lohar thereby making the recording and/or reconstruction process slow. We propose tosurmount this problem by recording a combined Fourier transform hologram andits computer reconstruction by inverse combined Fourier transformation. If g ( x, y ) is an object function, its forward Fourier transform (FT), i.e., F T { g ( x, y ) } ,is given by + ∞ (cid:90) −∞ + ∞ (cid:90) −∞ g ( x, y ) exp {− i π ( ux + vy ) } dxdy = G ( u, v ) (1)and its inverse FT, i.e., I F T { g ( x, y ) } , is given by + ∞ (cid:90) −∞ + ∞ (cid:90) −∞ g ( x, y ) exp { i π ( ux + vy ) } dxdy = G (cid:63) ( u, v ) (2)where x, y and u, v are the space and spatial frequency coordinates pairs, i = √− F T , I F T signify Fourier transform and imverse Fourier transform operator andthe symbol (cid:63) stands for complex conjugation. Following the definitions of FT andinverse FT, the identities below also hold I F T { g ( x, y ) } = G (cid:63) ( u, v ) = G ( − u, − v ) (3) I F T { g ( − x, − y ) } = G (cid:63) ( − u, − v ) = G ( u, v ) (4)We now define a combined Fourier transform (CFT) as [5] G cft ( u, v ) = a F T [ g ( x, y )] + a I F T [ g ( x, y )]= a G ( u, v ) + a G ( − u, − v ) (5)where a and a are two constant coefficients, real or complex, but must satisfy a (cid:54) = a . The object function g ( x, y ) can be recovered by an inverse CFT (ICFT)given by g ( x, y ) = ( a − a ) − [ a I F T { G cft ( u, v ) }− a F T { G cft ( u, v ) } ] (6) We record a CFT hologram by adding a coherent plane wave to the wave fielddistribution of equation (5). If R is the amplitude of the collimated referencewave, the wave field distribution at the hologram recording plane will be given by Ψ ( u, v ) = R + G cft ( u, v ) . (7)The recorded hologram intensity will be given by I ( u, v ) = | Ψ ( u, v ) | = | R | + | G cft | + RG (cid:63)cft + R (cid:63) G cft (8) win image elimination in digital holography 3 The amplitude of the reference wave R is made large enough so as to make thehologram recording linear. We also capture a separate record of the referencewave intensity | R | = R only, prior to recording the combined Fourier transformhologram. Subtracting R from both sides of equation (8) and also dividing bothsides by R , we get the modified hologram intensity as I (cid:48) = | G cft | R + 1 R { RG (cid:63)cft + R (cid:63) G cft } (9)Since, | R | (cid:29) | G cft | , | G cft | / | R | is vanishingly small, hence it can be neglectedand equation (9) reads I (cid:48) ≈ | R | { RG (cid:63)cft + R (cid:63) G cft } (10) The object function g ( x, y ) can be reconstructed from the CFT hologram by illu-minating the hologram by the reference wave R and by ICFT operation, i.e., I CF T [ RI (cid:48) ] = I CF T [ G (cid:63)cft ] + I CF T [ G cft ] (11)since R (cid:63) /R = 1 because R (cid:63) = R , and I CF T stands for ICFT operator. Using theidentities of equations (3) and (4) and expanding the inverse CFTs, equation (11)can be expressed as I CF T [ RI (cid:48) ] = A g ( x, y ) + A g ( − x, − y ) (12)where A and A are constants involving the weighting coefficients a and a givenby A = (cid:18) a a (cid:63) − a (cid:63) a | a | − | a | (cid:19) , A = (cid:18) a a (cid:63) − a a (cid:63) | a | − | a | (cid:19) (13)If we take values of a and a such that they are mutually complex conjugate, asfor example, if a = b + ib and a = b − ib (14) b and b being real, A = 1 but A = 0, and equation (12) reads (cid:8) I CF T [ RI (cid:48) ] (cid:9) a = a (cid:63) = g ( x, y ) (15)That means, only the object function g ( x, y ) is reconstructed and the conjugatereconstruction g ( − x, − y ) is eliminated. If a and a are not exactly complex con-jugates of each other, the twin image will be present. The strengths of the twoimages will depend on the real and imaginary parts of a and a .By De Moivre’s theorem, equation (14) can be expressed as a = B exp( iφ ) and a = B exp( − iφ ) (16)where B = (cid:113) b + b and φ = tan − ( b /b ) (17) Putting a and a , equation (5) becomes G cft ( u, v ) = B { exp( iφ ) G ( u, v )+ exp( − iφ ) G ( − u, − v ) } = B exp( − iφ ) { exp( i φ ) G ( u, v ) + G ( − u, − v ) } (18) Debesh Choudhury, Gautam Lohar
Now, we can create a wave field distribution that is equivalent to equation (18)using an optical arrangement of Fig.1 and record optical CFT holograms. An ex- g(x,y) g(-x,-y)FL FL M M FL QWPP BS M BS CCD
CollimatedLaser Beam P π/4 PBSQWP P θ Axes of PolarizingComponents P π/4 HWP α HWP P P θ PBS
Fig. 1
Recording geometry for combined FT holograms. The polarizing components and theirorientations are shown in the top left corner in a dashed box. panded collimated beam of light from a He-Ne laser (red colored online) is splitthrough a beam-splitter BS . One part is directed to a polarizing beam-splitterPBS. The two mirrors M and M form a triangular path interferometer, andthe transmitted/reflected beams recombine by PBS. The transmissive object withamplitude transmittance g ( x, y ) is placed inside the interferometer. We place oneFourier lens pair FL and FL of equal focal lengths inside the interferometer such that an inverted image of the object is formed as shown in Fig.1. The interfer-ometric arrangement is so adjusted such that the object g ( x, y ) and its invertedimage g ( − x, − y ) are equidistant from the beam-splitter PBS. So, one can get atthe output side of the interferometer Fresnel propagated wave fields from (i) theobject function g ( x, y ) and from (ii) an inverted version of the object function, win image elimination in digital holography 5 i.e., g ( − x, − y ). A third Fourier lens FL is also placed which produces the Fouriertransforms of g ( x, y ) and g ( − x, − y ) with appropriate coefficients that depend onthe phase relationship between the wave fields propagating through the interfero-metric arms. The orientations of the polarizing components are shown in a dashedbox. Another part of the input beam from the laser, which is transmitted fromthe beam-splitter BS and is reflected by the mirror M , superposes with objectbearing beams after transmitting through another beam-splitter BS . To get aCFT hologram at the CCD plane, the phase differences between the beams areadjusted to 2 φ using polarization-induced phase [6].The input laser beam is polarized by the polarizer P at angle π/ x -axis,and may be represented by a Jones vector ˆ E as [7]ˆ E = E (cid:20) (cid:21) (19) E being the amplitude. This polarized beam is split by the polarizing beam-splitter PBS into two orthognally polarized beam along the x -axis and y -axis, andthe Jones matrices corresponding to the PBS along the x -axis and y -axis may begiven by [7] P x = (cid:20) (cid:21) , P y = (cid:20) (cid:21) (20)The reflected polarized beam, after a round trip through the cyclic interferom-eter gets reflected by the PBS and comes out at the output side. Similarly, thetransmitted polarized component comes out of the PBS by transmission at theoutput side. It is to be noted that these two polarized beams carry the objecttransmittance information g ( x, y ) and its inversion g ( − x, − y ) by appropriate lenstransformation inside the interferometer. These two orthogonally polarized imagebearing beams face a quarter-wave retardation plate QWP whose slow axis makesan angle π/ x -axis. Finally, these two polarized beams after passingthrough the quarter-wave plate will pass through an analyzer P and the wavefields proceed towards the CCD camera. After Fourier transformation by the thirdlens, the vector wave field distribution at the CCD camera plane will be given byˆ G P I ( u, v ) = P ( θ )[ C ( π/ P x ˆ EG ( u, v )+ C ( π/ P y ˆ EG ( − u, − v )] (21)where P ( θ ) represents the Jones matrix of the analyzer P , C ( π/
4) represents theJones matrix of the quarter-wave retardation plate, and are given by [7] P ( θ ) = (cid:20) cos θ cos θ sin θ sin θ cos θ sin θ (cid:21) (22) C ( π/
4) = 12 (cid:20) i − i − i i (cid:21) (23) Performing matrix multiplications, equation (21) can be expressed asˆ G P I ( u, v ) = (cid:104) exp (cid:110) i (cid:16) π − θ (cid:17)(cid:111) G ( u, v ) + exp (cid:110) − i (cid:16) π − θ (cid:17)(cid:111) G ( − u, − v ) (cid:105) × (cid:20) cos θ sin θ (cid:21) (24) Debesh Choudhury, Gautam Lohar where we have dropped off the constant amplitude factors. If we put θ = π/ − φ ,the polarization-induced phase difference becomes 2 φ , and equation (24) becomesequivalent to the field distribution given by equation (18) and is nothing but theCFT of the object function g ( x, y ) except that it is linearly polarized. We can alsoidentify the phase factors exp { i ( π/ − θ ) } and exp {− i ( π/ − θ ) } as the complexcoefficients a and a respectively.Finally, an analyzer P should be placed before the CCD camera whose trans-mission axis makes an angle θ (same as P ) with the x -axis. A half-wave plateHWP should also be placed after the Fourier lens FL with its slow axis makingan angle α ( α (cid:54) = θ ) with the x -axis, so as to keeping the amplitudes of the objectwaves reaching the CCD camera much smaller than the amplitude of the referencewave. This will help to ensure that | R | (cid:29) | G cft | . The object waves and the ref-erence wave will transmit through the analyzer P and interfere to form thd CFThologram which can be recorded by the CCD camera. Although the derivations of the earlier sections are shown for continuous combinedFourier transform, it can be shown that the results are valid for discrete combinedFourier transform as well. We have carried out proof-of-the-principle study usingcomputer simulations by using GNU Octave [8]. The object transparency is of200x200 pixels size as shown in Fig.2(a). The simulation window size is 256x256pixels where the 200x200 pixels object is placed at the right hand top cornerwith zero padding. The object is purposefully off-centered so as to distinguishthe twin reconstructions. The computational reconstruction of the object froma discrete FT hologram is carried out and is shown in Fig.2(b). Simulations forobject reconstructions from the discrete CFT hologram can be done for two cases:Case I considering real object, and Case II considering complex object.6.1 Case I: Reconstruction with a real objectThis case is straight forward. Computational reconstructions from simulated dis-crete CFT holograms for different values of a and a are shown in Fig.2(c) and2(d) for a ≈ a (cid:63) ( b = 1 . b = 1) and for a = a (cid:63) ( b = b = 1) respectively. Itis evident from Fig.2(b) that reconstruction of the FT hologram reproduces twinimage with equal strengths. On the otherhand, for the CFT hologram, the effectof twin image is present for a (cid:54) = a (cid:63) ( b = 1 . b = 1), but the intensity levels ofthe two images are unequal [Fig.2(c)]. The effect of twin image is completely elim-inated for a = a (cid:63) ( b = b = 1) as discernible from Fig.2(d). It is clear from theresults of Fig.2(c) and Fig.2(d) that the effects of twin image in reconstructionsof CFT holograms can be controlled by proper choice of a and a , and can becompletely eliminated for a = a (cid:63) . win image elimination in digital holography 7(a) (b)(c) (d) Fig. 2
Simulation results with a real object: (a) Object; Reconstruction from (b) FT hologram,(c) CFT hologram with a ≈ a (cid:63) , and (d) CFT hologram with a = a (cid:63) . variation in thickness or refractive index of the material or both. If t ( x, y ) representsthe thickness distribution over the object transparency, the phase distribution dueto this thickness variation will be given by δ ( x, y ) = n (cid:18) πλ (cid:19) t ( x, y ) (25)where λ is the wavelength of the laser and n is the refractive index of the materialof the transparency. Here, it is assumed that the refractive index of the materialof the object transparency is uniform. We consider an example of a quadratic thickness variation, so t ( x, y ) may be expressed as t ( x, y ) = t + ( x + y ) ∆ (26)where t is the constant nominal thickness of the transparency plate and ∆ is athickness that is varied quadratically with spatial coordinates x, y . Therefore, the Debesh Choudhury, Gautam Lohar effect will be a phase factor exp[ iδ ( x.y )] multiplied to the object function. If weput ∆ = 0, we get a constant phase factor which is equivalent to the real objectof Case I.Simulations have been carried out with a 200x200 size phase function multipliedwith the 200x200 size object function for λ = 633 nm, t = 2 mm and differentvalues of ∆ , the overall simulation window remaining 256x256 size. The surfaceplot of the normalized phase function t ( x, y ) for ∆ = 0, and ∆ = 0 . λ are shownin Fig.3(a) and 3(b). The phase function profile is constant in Fig.3(a) which isnothing but the case of real object of Case I. The computational reconstructionsof CFT holograms are shown in Fig.3(c), 3(d), 3(e) and 3(f) for ∆ = 0, 0.5 λ , λ and 2 λ respectively. Here, the weighting multipliers are kept at a = a (cid:63) and b = b = 1 for all. It is evident from Fig.3(c) – 3(f) that the phase variation of theobject transparency gives rise to circular fringes modulated over the reconstructedobject image. The effect is negligible for ∆ = 0 . λ [Fig.3(d)], but as ∆ increasesto λ , a circular fringe appears near the periphery of the image [Fig.3(e)]. When ∆ = 2 λ , the image is modulated by two circular fringes [Fig.3(f)]. It is clear thatthe thickness of the object transparency should be optically uniform, i.e., thesurface finish of the object transparency plays an important role. That is whyin such an interferometric arrangement, the object transparency may be placedinside an optical tank filled with an index matching liquid and optically polishedglass windows with surface finish better than λ/ The proof-of-the-principle simulations of the previous section proved the feasibilityof twin image elimination in digital in-line holography using CFT. The proposedsystem is a polarization triangular path interferometer with a couple of Fourierlenses for synthesizing an inversion of the object function [ g ( − x, − y )], along withthe object function [ g ( x, y )], and another Fourier lens for performing Fourier trans-form of the object and its inversion. The effect of CFT is created by introducingappropriate phase difference between the object information bearing waves by apolarization technique. Simulation is carried out for generating the CFT hologramsfor different values of the complex weighting factors. In this simulation, we havenot considered the practical aspects of experimentation, such as effects of imper-fections of the polarization components, aberrations of the Fourier lenses and themirrors, misalignment errors of the interferometer, object wave to reference waveratios and the quantization errors of the CCD camera.The alignment of the proposed interferomeric system is not easy, neverthe-less it is possible to implement combined Fourier transform optically. From equa-tions (16), (17) and (18) it is evident that the complex constants a and a aretransformed into a phase difference 2 φ . So, in optical implementation, the constants a and a are not required to be specified directly, instead the orientations of thepolarizing components are so adjusted such that the phase difference 2 θ between the two image bearing object waves can be made equal to 2 φ . If one desires, onecan easily express the induced phase differences equivalent to the complex factors a and a using De Moivre’s relations of equations (16) and (17). The complexcoefficients a and a are also identified in the output of the interferometer inequation (24) as exponential phase factors. win image elimination in digital holography 9(a) (b)(c) (d)(e) (f) Fig. 3
Simulation results with a complex object: The normalized phase function t ( x, y ) isshown for (a) ∆ = 0 and (b) ∆ = 0 . λ . Reconstructions from the CFT hologram are shownfor (c) ∆ = 0, (d) ∆ = 0 . λ , (e) ∆ = λ and (f) ∆ = 2 λ .0 Debesh Choudhury, Gautam Lohar Since, it is an interferometric transformation, the holographic reconstructionis sensitive to phase aberrations of the object transparency. Thus, this techniquemay be useful for carrying out studies of phase objects. A similar inverting trian-gular path polarization interferometer was indeed utilized for implementing opticalHartley transform experimentally [9]. We wish to consider the practical aspects ina future experimental implemetation of the proposed method.
We have proposed a technique for resolving twin image problem of in-line digitalFourier holography. We use a combination of Fourier transforms to eliminate theunwanted twin image. The technique relies on interferometric addition of Fouriertransforms of the object and its inversion with appropriate complex weightingcoefficients, which are introduced through polarization-induced phase. Simulationresults prove the feasibility of complete elimination of one of the twin image. More-over, the technique is sensitive to phase aberrations of the object transparency,thereby providing a way for phase sensing holographic imaging. The proposedmethod neither involves recording of several phase shifted holograms, nor it re-quires any iteration for digital reconstruction, which may render it suitable forrecording holograms of fast changing sequences as well as fast digital reconstruc-tion of the recorded holograms.
Acknowledgements
The authors thank an anonymous reviewer for fruitful comments thathelped to improve the paper.
References
1. D. Gabor, “A new microscopic principle,” Nature , 777-778 (1948).2. W. L. Bragg and G. L. Rogers, “Elimination of the unwanted image in diffraction mi-croscopy,” Nature , 190 (1951).3. B. M. Hennelly, D. P. Kelly, N. Pandey, D. Monaghan, “Review of twin reduction andtwin removal techniques in holography,” Proc. China-Ireland International Conference onInformation and Communications Technologies, Maynooth, Ireland, Pages 241-245 (2009).4. E. Leith and J. Upatnieks, “Wavefront reconstruction with continuous-tone objects,”JOSA CAS-32 (6), 618-619 (1985).6. D. Choudhury, P. N. Puntambekar and A. K. Chakraborty, “Utilization of polarization-induced phase difference for complex addition and subtraction of amplitudes,” J. Opt. ,6-10 (1993).7. A. Gerald and J.M. Burch, Introduction to Matrix Methods in Optics26