Viewing angle enhancement for two- and three-dimensional holographic displays with random superresolution phase masks

Edward Buckley, Adrian Cable, Nic Lawrence, and Tim Wilkinson

Holographic displays employing binary phase modulation have been demonstrated to be attractive on the grounds of efciency and miniaturization, and they offer a plausible approach to two-dimensional (2D) and three-dimensional (3D) image projection and display. A novel algorithm one-step phase retrieval and corresponding hardware architecture have recently been proposed, providing the performance required for real-time holographic display. However, since viewing angle varies inversely with pixel size, very small display pixels are required to achieve a wide eld of view. This is particularly problematic for 3D displays, as the requirement for a large display with small pixels has hitherto necessitated an unachievably large electrical bandwidth. We present a novel approach, utilizing xed random pixelated quaternary phase masks of greater resolution than the displayed hologram, to dramatically increase the viewing angle for 2D and 3D holographic displays without incurring a bandwidth penalty or signicantly degrading image quality. Furthermore, an algorithm is presented to generate holograms accounting for the presence of such a phase mask, so that only one mask is required. 2006 Optical Society of America OCIS codes: 090.1760, 050.1970.

1. Introduction

Computer-generated binary phase holograms have long been considered an attractive basis for twodimensional (2D) and three-dimensional (3D) information display because of the efciency, robustness, and potential for miniaturization. While an appropriate device for the rapid dynamic display of such hologramsthe ferroelectric liquid-crystal (FLC) spatial light modulator (SLM) has been available for a long time, the computational demands of the hologram generation algorithms have until recently precluded the development of a real-time holographic video display. However, the advent of the one-step phase retrieval (OSPR) algorithm1,2 and advances in eld-programmable gate array and applicationspecic integrated circuit technology have at last brought such a display within reach.

E. Buckley (ebuckley@iee.org), A. Cable, and N. Lawrence are with Light Blue Optics, Limited, St. Johns Innovation Centre, Cowley Road, Cambridge CB4 0WS, UK. T. Wilkinson is with the Department of Engineering, Cambridge University, Trumpington Street, Cambridge CB2 1PZ, UK. Received 28 March 2006; revised 21 May 2006; accepted 21 May 2006; posted 30 May 2006 (Doc. ID 69419). 0003-6935/06/287334-08$15.00/0 2006 Optical Society of America 7334 APPLIED OPTICS Vol. 45, No. 28 1 October 2006

However, there are still a number of issues that should be addressed before such a display can offer a subjective level of performance to compete with conventional 2D technologies, or offer 3D image display for general applications. It is well known that any diffractive element (such as a binary phase hologram) that imparts a purely real modulation on an incident wavefront produces a conjugate image in the replay eld (RPF), thereby reducing the usable display area and optical efciency by half.3 Although it is straightforward to generate quaternary phase holograms that would solve this problem, such holograms cannot be displayed on the inherently binary ferroelectric devices available. A second problem arises from the fact that the viewing angle of a holographic display varies inversely with pixel size. As a result, small display pixels are required to achieve a wide eld of view, and so an inordinate amount of bandwidth is required to display a large image. In this paper we demonstrate that one binary pixelated phase mask of resolution M M, identical in physical size to that of the P P pixel SLM but of greater resolution, can be used to increase the viewing angle of the overall system by a factor of approximately F PM. Furthermore, this can readily be combined with the phase mask required for conjugate image removal to form a quaternary phase mask that we show can be used to ameliorate the two aforemen-

Fig. 1. Effective phase pattern formed by hologram and phase-mask pixels when aligned and placed in close contact: (a) hologram pattern huv, (b) phase-mask pattern puv, (c) effective pattern puvhuv.

tioned problems with present binary holographic display systems in one step. We show that, for a given number of on points in the RPF, increasing the viewing angle by a factor of 2 incurs a signal-to-noise ratio (SNR) penalty of approximately 3 dB. However, further increases in viewing angle by increasing the phase-mask resolution results in a negligible additional SNR drop. The same technique allows an increase in the number of addressable points in the RPF from P P to M M; although this improvement comes at the expense of additional RPF noise compared with a true M M pixel hologram, the perceptual effect is automatically attenuated through use of the OSPR algorithm, which reduces RPF noise variance by displaying multiple subholograms per frame and exploiting ocular temporal integration. Hence the use of the superresolution phase mask in conjunction with OSPR to generate multiple subholograms per frame effectively allows the exchange of spatial for temporal bandwidth in a holographic display.
2. Conjugate Image Removal Using a Binary Phase Mask

It was rst suggested in the context of optical correlators4 that the conjugate image manifest in the RPF, caused by the purely real 0, modulation imparted by a binary phase ferroelectric SLM, could be suppressed by using a spatially random binary pixelated phase mask. The principle of operation is described in Fig. 1. The phase mask puv, which consists of a pixelated pattern of the same pitch as the hologram pixels, is placed in close contact and aligned to the hologram pattern huv displayed on an SLM. If the phase-mask pixels impart phase modulation in the set 0, 2 rad and the SLM pixels retard the incident light by 0, rad, then the effective phase pattern of Fig. 1(c) results, with the pixels imparting a net modulation in the set 0, 2, , 32. This provides the extra degree of freedom required to suppress the conjugate image, despite the fact that the SLM itself is binary. Some example simulated results are shown in Fig. 2. A conjugate-symmetric image Fxy is shown in Fig. 2(a), which results from the replay of a binary phase hologram huv where Fxy huv and is the Fourier transform. In Fig. 2(a), however, a hologram has been designed that accounts for the presence of the phase mask puv, giving rise to a RPF Fxy

huvpuv that is devoid of the conjugate image. This comes at the expense, however, of additional RPF noise. We performed a series of simulations to investigate the effect of hologram resolution and number of on points in the target RPF on the SNR, with and without a phase mask. It was found that, for a given number of on points (500 points with random spatial distribution), doubling the resolution in both directions increases the SNR by around 6 dB. Furthermore, for a given hologram resolution, doubling the number of on points decreases the SNR by 3 dB. In both cases, the utilization of a phase mask can be seen to cause an additional degradation in SNR of approximately 3 dB. These relationships are shown in Fig. 3. We have previously demonstrated the OSPR algorithm, which exploits the large temporal bandwidth of a ferroelectric liquid-crystal-on-silica SLM to ameliorate the perceptual signicance of the constant SNR degradation caused by the presence of the phase mask. Hence the use of a phase mask to remove the conjugate image, thereby doubling the number of addressable points in the RPF, is well suited to video display applications. This result motivates the investigation of the possibility of applying this phasemasking technique to ameliorate the other major problem with holographic display, namely, that of limited viewing angle.
3. Viewing Angle Increase Using a Superresolution Phase Mask

As noted in Section 2, the inverse relationship between viewing angle and pixel size means that, to

Fig. 2. RPFs resulting from (a) a binary phase hologram huv and (b) a phase mask and hologram combination huvpuv. 1 October 2006 Vol. 45, No. 28 APPLIED OPTICS 7335

Fig. 3. Effect of (a) hologram resolution and (b) number of on pixels upon RPF SNR, with and without a phase mask.

achieve a display with a wide eld of view, a hologram with a small feature size is required. More specically, the viewing angle of a hologram illuminated by coherent light of wavelength varies inversely with pixel size according to the equation 2 arctan . 2 (1)

It is therefore desirable to reduce the pixel size as far as possible, since if the pixels are sufciently small, then subsequent demagnication optics are effectively obviated. For example, at a wavelength of 532 nm, 1.2 m pixels would give rise to a viewing angle of 30 deg and a RPF size of 0.44 m 0.44 m at 1 m from the SLM. For a hologram of xed size this can clearly be achieved by increasing the resolution, making each pixel smaller. However, this becomes increasingly difcult to achieve for dynamically addressable devices due to electrical bandwidth, materials, and fabrication constraints. Indeed ferroelectric liquidcrystal-on-silicon devices, which switch sufciently fast to allow display of multiple holographic subframes per image frame for the purposes of noise reduction, are fundamentally compromised at small pixel sizes due to the limitations imposed by the use of FLC material in conjunction with complementary metal-oxide semiconductor backplane technology. To maintain both high ll factor and small pixels, the
7336 APPLIED OPTICS Vol. 45, No. 28 1 October 2006

switching voltage must scale with respect to pixel size to avoid eld breakdown across the silicon backplane. However, to switch sufciently rapidly, FLC material requires relatively large switching voltages (approximately 15 V) and it is therefore difcult to reduce the pixel size below 10 m. Even if new low-voltage FLC materials became available, a large number of small display pixels would require an inordinate electrical bandwidth since, for the same pixel geometry, a display just 20 mm 20 mm in size would require approximately 3 108 pixels. Assuming one hologram pattern per monochrome video frame, then using this device to display a frame-sequential full color video stream at 25 framess would require a sustained data transfer rate of around 23 Gbitsswell beyond current technology both in terms of storage and electrical bus design. The alternative approach, presented here, is to extend the concept of the phase mask. The phase-mask technique is extended to demonstrate that a binary 0, phase mask prs of resolution M M, identical in physical size to that of the P P pixel SLM but of greater resolution, can be used to increase the viewing angle of the overall system by a factor of approximately PM. This technique increases the number of addressable points in the RPF from P P to M M, at the cost of introducing additional RPF noise, which reduces the RPF SNR by 3 dB. However, this SNR drop is virtually independent of the viewing angle increase, and, although the RPF noise is greater compared with a true M M pixel hologram, the perceptual effect can be automatically attenuated by the use of multiple OSPR subframes. Hence the use of the superresolution phase mask in conjunction with OSPR, generating multiple subholograms per frame to reduce noise variance, effectively allows the exchange of spatial for temporal bandwidth in a holographic display. This approach relaxes the demands imposed upon the display technology required for high-quality 2D and 3D holographic projection. The OSPR algorithm has previously been presented as a novel method of hologram generation and display, exploiting the human perception of statistical image noise parameters encountered in holographically generated images to generate images of substantially improved quality compared with other commonly utilized algorithms such as direct binary search. The algorithm begins with the specication of a P P pixel target intensity image Txy, returning a set of N individual P P pixel holograms huvn. Each of the noise elds produced by the N holograms are independent identically distributed (i.i.d.), and hence when N holograms are displayed within the integration period of the eye, ocular temporal averaging occurs and the noise variance of the resultant RPFs fall as 1N. A modied version of the OSPR algorithm was derived to account for the presence of a phase mask and is detailed in Algorithm 1 in Appendix A. Given an M M target image Txy, the algorithm accounts for the presence of the superresolution M M pixel

Fig. 4. Simulated RPFs produced by 256 256 holograms (a) without and (b) with a 512 512 phase mask. Fig. 5. SNR variation with a resolution of 0, phase mask employed in conjunction with 256 256 pixel holograms.

phase mask puv to generate 2N distinct P P pixel holograms huvn where P is an integer multiple of M. When the phase mask and hologram are then placed in close contact and illuminated by coherent light, the resulting RPF Fxy puvhuv is at the higher resolution of M M, thereby overcoming the conventional constraint whereby a P P pixel hologram cannot be designed to form more than P P points in the RPF. This modied algorithm also includes an improvement to generate two holograms exhibiting i.i.d. RPF noise per Fourier transform, in comparison with the original algorithm that produced just one. Step 1 of Algorithm 1 forms N targets equal to the amplitude of the supplied intensity target Txy, but with i.i.d. uniformly random phase. Step 2 computes the inverse Fourier transform, taking into account the presence of the phase mask puv, to produce continuous complex holograms of size M M. Step 3 averages the resultant complex hologram over blocks of size F F to produce suvn, a set of N averaged complex hologram elds of the required size P P. Two independent holograms are then generated in steps 4 and 5 from each single complex hologram eld suvn. Binarization of these holograms is then performed in step 6, as per previous implementations of the OSPR algorithm; thresholding around the median of muvn ensures that equal numbers of 1 and 1 points are present in the holograms, achieving dc balance (by denition) and also minimal reconstruction error.5 Details of the derivation of the new algorithm are given in Appendix A.
4. Binary Phase Mask for Viewing Angle Increase

To demonstrate the efcacy of the modied OSPR algorithm, a target image was selected and single binary 0, 256 256 holograms were generated for use with and without a binary 0, 512 512 phase mask. The simulated RPFs both of which have the same physical dimensionsare shown in Fig. 4. From Fig. 4(a) it can be seen that the viewing angle is limited by the presence of side orders in the RPF. The area inside the rectangle is the rst-order diffraction pattern determined by the pixel pitch of the hologram; outside the rectangle are the repeated higher orders, which cannot be controlled by the ho-

logram alone. When the phase mask is employed as shown in Fig. 4(b), the addressable RPF area, and hence the viewing angle, is doubled in each dimension. However, a drop in SNR is observed due to the presence of the superresolution phase mask. We have already shown how SNR is affected by the number of on points in the RPF and by hologram resolution when no phase mask is present. In Fig. 5 we proceed to show how the SNR is also affected by the phasemask resolution for a 256 256 OSPR-generated hologram N 1 forming a 500-point RPF. Although a signicant decrease in SNR of 3.6 dB results from the doubling of the viewing angle using a 512 512 phase mask, we nd surprisingly that further increases in the viewing angle result in a negligible subsequent decrease in SNR. For example, increasing the viewing angle by 12 times (by using a 3072 3072 phase mask) results in an SNR drop of just 4.1 dB compared with no phase mask at all. This SNR degradation of an image reconstructed from a hologram and 0, phase-mask combination can be characterized by examining three different cases corresponding to no viewing angle increase, where the hologram and phase mask are the same resolution F 1 and there are viewing angle increases of two and greater F 2. When the hologram and phase-mask resolutions are equal, then F 1 and there is no increase in the viewing angle. If the hologram is designed using Algorithm 1, the hologram pixels will be chosen to account for the presence of the 0, phase mask and hence there is no fundamental difference in the replay of a hologram and the replay of a hologram and 0, phase-mask combination of the same resolution. Hence, any RPF noise is entirely due to the binarization of the hologram. When F 2, however, the viewing angle is doubled and twice as many pixels in the RPF can be addressed. As a consequence, the noise energy is doubled compared with the no phasemask case and a 3 dB reduction in SNR results. A further doubling of the viewing angle would require a fourfold increase in the phase-mask resolution, causing an additional doubling of noise energy. However, the noise energy increase is accompanied by a further
1 October 2006 Vol. 45, No. 28 APPLIED OPTICS 7337

Fig. 6. (a) Simulated and (b) measured performance of a 512 512 pixel hologram used with a 1024 1024 pixel phase mask to double the viewing angle. Fig. 7. SNR degradation caused by use of different types of phase mask. Case 1, no phase mask; case 2, conjugate image removal using a 0, 2 phase mask; case 3, 0, superresolution phase mask used to double the viewing angle; case 4, conjugate image removal and viewing angle doubling using a quaternary 0, 2, , 32 phase mask.

quadrupling of the RPF addressable area so that the additional SNR degradation is much less than 3 dB. To demonstrate the operation of the superresolution phase-mask technique, binary phase diffractive optical elements (DOEs) were made to simulate the operation of a SLM and phase mask. A 512 512 pixel hologram DOE was calculated using the OSPR algorithm with N 1, accounting for the presence of a spatially random binary phase mask. The pixel size was 40 m, so that the active area was approximately 20 mm. The spatially random phase-mask DOE was designed to contain 1024 1024 pixels, each of size 20 m, resulting in a viewing angle increase of F 2. Both DOEs were made by e-beam etching of a 1.5 mm thick fused-silica substrate of refractive index 1.46, with a step height of 0.65 m to optimize operation at a wavelength of 532 nm. The phase-mask and hologram DOEs were placed in close contact and carefully aligned using custom optical mounts so that the pixels were accurately registered. The elements were illuminated using a 10 mW green laser coupled into a single-mode ber, and since the diffraction angle from the hologram and phase-mask combination was small, an objective lens of focal length 100 mm was used to image the resultant RPF onto a complementary metal-oxide semiconductor sensor array of active area 25 mm 25 mm. The simulated and captured RPFs are shown in Figs. 6(a) and 6(b), respectively; the AJC EB image corresponds to the rst-order diffraction pattern that would occur from a hologram with pixel size 40 m, whereas the EXTRA pattern is due to the increased viewing angle provided by the 20 m pixels of the phase mask. If a 512 512 pixel hologram had been used alone, the space occupied by the EXTRA image would have contained the overlapping secondary orders of Fig. 4(a).
5. Quaternary Phase Mask for Viewing Angle Increase and Conjugate Image Removal

modulation in the set 0, . Utilization of each mask individually incurs a SNR penalty of approximately 3 dB. A reasonable conjecture is that the use of both a 0, 2 mask and a 0, mask in an optical system would facilitate simultaneous removal of the conjugate image and an increase in viewing angle. Since the optical system is linear, the total phase modulation imparted on the incident wave by both masks together is equal to the sum of the phase shifts imparted by each mask. As a result, we can use a single random quaternary phase mask whose phases lie in the set 0, 2, , 32. Surprisingly, combining these masks into a 0, 2, , 32 mask to both double the viewing angle and remove the conjugate image does not give the expected 6 dB penalty, but instead results in a SNR degradation of only 3.8 dB approximately. This general result is found to hold independent of resolution or on-pixel count, as is shown in Fig. 7. We nd, as established with the 0, mask, that further increases in viewing angle can be achieved at little expense in terms of RPF SNR degradation. This is demonstrated for a variety of hologram and phasemask resolutions in Fig. 8.

We have shown that successful removal of the conjugate image can be accomplished with a 0, 2 random phase mask and that the viewing angle can be increased through use of another random superresolution phase mask, albeit with pixels imparting phase
7338 APPLIED OPTICS Vol. 45, No. 28 1 October 2006

Fig. 8. SNR variation with various hologram and phase-mask resolutions for a RPF with 500 on pixels.

Fig. 9. Simulated RPFs produced by OSPR Algorithm 1 using a quaternary phase mask for different numbers of hologram subframes N.

6. Compensation for Additional Noise by One-Step Phase Retrieval

We have shown that use of the OSPR technique to increase the viewing angle and remove the conjugate image introduces a xed but signicant drop in SNR of around 3 dB. We propose that, as previously shown,1 the perceptual signicance of the RPF noise can be substantially attenuated by generating multiple subframes using the OSPR algorithm with N 1. Ocular temporal averaging is then exploited to yield a signicantly reduced perceived noise level, even though the SNR is not improved. We demonstrate in simulation the perceived decrease in RPF noise that results for different numbers of subframes N generated by the OSPR algorithm and subsequently displayed sequentially within a 125 s time interval. Figure 9 demonstrates that increasing the number of subframes N signicantly improves the perceived quality of the resultant RPF image, despite the fact that the SNR varies negligibly between each of the images.
7. Manufacture of the Quaternary Phase Mask

Although we have shown that the combination of a quaternary phase mask and a binary SLM can be utilized in conjunction with OSPR to produce a wide viewing angle, a conjugate image-free display with a RPF of high perceptual quality, we have so far not addressed the issues involved in manufacturing such a mask. Previous efforts in manufacturing random pixelated phase masks have proceeded by rst producing a premask as a binary amplitude pattern, either printed on plastic or as a chrome-on-glass structure. UV-sensitive photoresist is then spun onto glass and the required phase pattern is produced photolithographically using

the premask, exposing the photoresist for the necessary duration to form the required phase steps. However, production of a four-phase mask using this technique is complicated, rst because this requires variable amplitude levels on the premask and second because these amplitude levels need to be precisely calibrated with respect to the exposure process to ensure that the required depths of the phase step are obtained. However, we have determined that it is possible to produce a suitable four-phase mask using just a single random binary amplitude premask, which can be produced using conventional processes. We begin by determining the exposure times t and t2 necessary to form steps of and 2, respectively, which can be achieved by examining the surface prole using a scanning surface prolometer (Dektak) for different exposure times. We then spin photoresist onto glass as before and UV expose the mask for the duration t. Under red light, so as not to expose the photoresist further, we rotate the premask by 90 deg and realign it with the previously exposed pattern. Exposure for a further time of t2 produces the required fourphase mask. It has been shown in simulation that a mask produced in this way, while not truly spatially random in phase, exhibits a sufcient spatial decorrelation between pixels to ensure that the reconstruction error compared to a truly random quaternary phase is minimal.
8. Extension to Three-Dimensional Image Display

One of the merits of this approach is the ability, for a xed available display data bandwidth, to substantially increase the viewing angle in a holographic display without incurring substantial degradation in
1 October 2006 Vol. 45, No. 28 APPLIED OPTICS 7339

Fig. 10. Simulated RPFs generated from (a) 1024 1024 hologram with 4096 4096 random quaternary phase mask and (b) true 4096 4096 hologram.

image quality. Since a very limited viewing angle is one of the major problems with 3D holographic displays created thus far, we believe that this approach constitutes a signicant advance toward an effective wide viewing angle 3D holographic display that is feasible with current technology. We consider the problem of designing a 3D display of physical size 10 mm 10 mm providing a viewing angle of around 14 deg, which would require 4096 4096 pixels of 2.5 m in size, necessitating a data bandwidth of 16 Mbitsframe. Instead, we propose manufacturing a xed 10 mm 10 mm quaternary phase mask of 4096 4096 pixels and employing in conjunction a 10 mm 10 mm SLM of only 1024 1024 pixels, each of size 10 m. Such a display is readily available from ForthDD displays, part number SXGA-R2-H1. With this architecture, we preserve the viewing angle of 14 deg but now require a data bandwidth of only 1 Mbitframe. As a demonstration of the efcacy of this approach, we calculated both a single standard 4096 4096 hologram and also a single 1024 1024 hologram using a 3D variant of OSPR with N 1 and a 4096 4096 quaternary phase mask. The 3D scene used in both cases was a set of 944 point sources that formed a wire-frame cuboid close to the hologram plane. Small sections of the simulated RPFs reconstructed from each hologram, obtained by propagating Huygens wavelets through a pinhole aperture, are shown in Fig. 10. As can be seen, Fig. 10(a) is of comparable quality to Fig. 10(b), despite the 16-fold reduction in hologram bandwidth, as expected from the results of previous simulations. While Fig. 10(a) does display an increased noise level, this can, as before, be compensated for by using the necessary number of OSPR subframes.
9. Conclusion

A novel approach of increasing the viewing angle was presented, utilizing a superresolution xed pixelated phase mask that imparted 0, modulation. It was found that, although doubling the viewing angle incurred a 3 dB drop in contrast, the viewing angle could be increased arbitrarily with only a small further decrease in contrast. An increase in viewing angle of 12 times was demonstrated with a resultant SNR decrease of just 4.1 dB. Finally, it was demonstrated that these improvements could be combined into one quaternary phase mask, providing increased viewing angle and conjugate image suppression for a SNR penalty of approximately 4 dB, regardless of RPF content. The hologram-generating algorithm was able to account for the presence of such a mask, so that only one mask is required for all hologram patterns. This technology therefore provides a dramatic increase in viewing angle for 2D and 3D holographic displays without the need to construct SLMs with large numbers of very small pixels or huge electrical bandwidth requirements.
Appendix A. Derivation of Modied One-Step Phase Retrieval Algorithm

In this Appendix, we detail the modications to the original OSPR algorithm (see Algorithm 2) that enable two output holograms to be formed per Fourier-transform operation instead of one. We begin by quoting the original algorithm, which generates sets of N distinct M M binary phase holograms huv, each producing a RPF with i.i.d. noise that approximates the same target image. The algorithm begins with the specication of a target intensity image Txy and proceeds as follows. Step 1 forms N targets Txyn equal to the amplitude of the supplied intensity target Ixy, but with i.i.d. uniformly random phase. Steps 2 and 3 shift the inverse Fourier transform holograms by a large distance to the right in the complex plane. This has the effect of making the phase of each point in the holograms very small, so that when we take their magnitude in step 4 (forcing the phase of every point to zero), we introduce practically no error. However, we note that lim

guvn

uv

1;

(A1)

so for a large , we can replace steps 3 and 4 with muvn guvn. (A2)

We have demonstrated methods, using pixelated binary phase masks, of mitigating the two remaining issues of holographic displays, namely, the presence of the conjugate image when employing a real-valued SLM and the limited viewing angle. The conjugate image could be suppressed by employing a phase mask with pixels that imparted phase modulation in the set 0, 2, at the expense of a 3 dB drop in RPF SNR.
7340 APPLIED OPTICS Vol. 45, No. 28 1 October 2006

Binarization of the hologram is then performed in step 5 by thresholding about the median of muvn as before. However, thresholding guvn around its median is equivalent to thresholding just guvn around its median. The statistical independence between the real and the imaginary parts of the discrete Fourier transform can now be exploited to calculate a second hologram muvnN guvn and its binarization huvnN that,

being uncorrelated with the rst hologram huvn, generates a RPF whose noise is also independent from the rst. As a result, using just a single Fourier transform, a pair of binary holograms can be generated both of which exhibit i.i.d. noise in their respective reconstructions, satisfying the condition required for the central limit theorem to apply. Algorithm 1. Modied OSPR algorithm for the generation of P P pixel holograms, given the presence of a supplied M M pixel phase mask. When the phase mask and hologram are placed in close contact and illuminated by coherent light, the resulting RPF is at the higher resolution of M M. 1. Let Txy Ixy exp jxy , where xy is uniformly distributed between 0 and 2 and n 1, . . . , N, x, y 1, . . . , M. 1 1 2. Let guvn Txyn, where puv is the puv phase mask and u, v 1, . . . , M, with M P. n F F 3. Let suvn k 1 l1 gFuk1,Fvl1 , where F MP. 4. Let muvn suvn, for n 1, . . . , N2 and u, v 1, . . . , P. 5. Let muvnN2 suvn, for n 1, . . . , N2 and u, v 1, . . . , P. 6. Let
n n n

Algorithm 2. OSPR algorithm. 1. Let Txyn Ixy exp jxyn, where xyn is uniformly distributed between 0 and 2 and x, y 1, . . . , N. 2. Let guvn 1Txyn, where 1 represents the 2D inverse Fourier-transform operator. 3. Let R be the smallest positive real value such that |guvn| Ru, v, n. 4. Let muvn | guvn|, where with R. 5. Let 1 if muvn Q, where Q median muvn huvn . 1 if muvn Q

References
1. A. J. Cable, E. Buckley, P. Mash, N. A. Lawrence, T. D. Wilkinson, and W. A. Crossland, Real-time binary hologram generation for high-quality video projection applications, in SID 04 Digest, (Society for Information Display, 2004), Vol. 53.1, pp. 13. 2. E. Buckley and A. J. Cable, Holographic apparatus and method, UK patent application GB0329012.9, P36148WO (13 December 2004). 3. A. G. Kirk and T. J. Hall, Design of binary computer-generated holograms: coding density and reconstruction error, Opt. Commun. 94, 491 496 (1992). 4. T. D. Wilkinson, D. C. OBrien, and R. J. Mears, Dynamic asymmetric binary holograms using a ferroelectric liquid crystal spatial light modulator, Opt. Commun. 109, 222226 (1994). 5. N. C. Gallagher, Jr., Optimum quantization in digital holography, Appl. Opt. 17, 109 115 (1978).

huvn

1 1

if muvn Q, where Q median muvn if muvn Q, and n 1, . . . , N

1 October 2006 Vol. 45, No. 28 APPLIED OPTICS

7341