Tiled large-screen three-dimensional display consisting of frameless multi-view display modules Yasuhiro Takaki,* Masayuki Tokoro, and Kenji Hirabayashi Institute of Engineering, Tokyo University of Agriculture and Technology, 2-24-16 Naka-cho, Koganei, Tokyo 1848588, Japan *
[email protected]
Abstract: To realize large-screen three-dimensional (3D) displays, frameless multi-view display modules are arranged two-dimensionally. This paper proposes a multi-view display module in which a multi-view flatpanel display is projected onto a screen of the module to provide a frameless screen. The display module consists of a multi-view flat-panel display, an imaging lens, an aperture, a screen lens, and a vertical diffuser. Prototype display modules were constructed having a screen size of 27.3 in., a 3D resolution of 320 × 200, and 144 viewpoints. Four modules were tiled vertically to provide a screen size of 62.4 in. Distortions in the screen imaging and viewpoint generation were corrected. ©2014 Optical Society of America OCIS codes: (110.0110) Imaging systems; (120.2040) Displays.
References and links 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
T. Okoshi, Three-Dimensional Imaging Techniques (Academic, 1976). N. A. Dodgson, “Autostereoscopic 3D displays,” Computer 38(8), 31–36 (2005). J.-Y. Son and B. Javidi, “Three-dimensional imaging methods based on multiview images,” J. Disp. Technol. 1(1), 125–140 (2005). H. Urey, K. Chellappan, E. Erden, and P. Surman, “State of the art in stereoscopic and autostereoscopic displays,” Proc. IEEE 99(4), 540–555 (2011). S. Iwasawa, M. Kawakita, S. Yano, and H. Ando, “Implementation of autostereoscopic HD projection display with dense horizontal parallax,” Proc. SPIE 7863, 78630T (2011). T. Balogh, “The HoloVizio system,” Proc. SPIE 6055, 60550U (2006). J.-H. Lee, J. Park, D. Nam, S. Y. Choi, D.-S. Park, and C. Y. Kim, “Optimal projector configuration design for 300-Mpixel multi-projection 3D display,” Opt. Express 21(22), 26820–26835 (2013). W.-L. Chen, C.-H. Tsai, C.-S. Wu, C.-Y. Chen, and S.-C. Cheng, “A high-resolution autostereoscopic display system with a wide viewing angle using an LCOS projector array,” J. Soc. Inf. Disp. 18(9), 647–653 (2010). W.-L. Chen, H.-H. Huang, T. H. Hsu, M.-H. Kuo, and C.-H. Tsai, “Optical simulation for cross-talk evaluation and improvement of autostereoscopic 3-D displays with a projector array,” J. Soc. Inf. Disp. 18(9), 662–667 (2010). T. Hashiba and Y. Takaki, “Development of a 3D pixel module for an ultra large screen 3D display,” Proc. SPIE 5599, 24–31 (2004). Y. Takaki, H. Takenaka, Y. Morimoto, O. Konuma, and K. Hirabayashi, “Multi-view display module employing MEMS projector array,” Opt. Express 20(27), 28257–28266 (2012). R. Kooima, A. Prudhomme, J. Schulze, D. Sandin, and T. DeFanti, “A multi-viewer tiled autostereoscopic virtual reality display, ” in Proceedings of the 17th ACM Symposium on Virtual Reality Software and Technology (Hong Kong, 2010), pp.171–174. S. Iwasawa, M. Kawkita, and N. Inoue, “REI: an automultiscope projection display,” Proceedings of 3D System and Applications, (Osaka, Japan, 2013), selected paper 1. M. Yamasaki, H. Sakai, T. Koike, and M. Oikawa, “Full-parallax autostereoscopic display with scalable lateral resolution using overlaid multiple projection,” J. Soc. Inf. Disp. 18(7), 494–500 (2010).
1. Introduction One potential application for large-screen and glasses-free three-dimensional (3D) displays is life-size, realistic communication. For this purpose, the screen should be large enough to fully cover one wall of a room, with the ability to support multiple viewers. In this study, to realize inexpensive and small-space implementation, easy installation and relocation, and flexible
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6210
screen configuration, we propose the tiling of frameless multi-view display modules that employ flat-panel displays. The multi-view display technique [1–4], which generates multiple viewpoints for observing corresponding parallax images, is usually used to construct large-screen and glasses-free 3D displays. These multi-view displays are constructed using systems such as multi-projection and flat-panel systems. The multi-projection system [5–7] is most commonly used to construct large-screen 3D displays that consist of an array of projectors and a common screen. In Ref. 5, all images generated by multiple projectors were superimposed on a common screen by properly shifting the projector lenses. A system that consisted of 200 projectors and a 200-in. screen was constructed. The projection length was 8.0 m because a long projection length is required to obtain a large screen. In Ref. 6, two mirrors were placed at the left and right sides of the multi-projection system, in order to increase the viewing angle. A system with a 140-in. screen and projection length of 5.6 m was commercialized using this technique. In Ref. 7, a display system that used two side mirrors was also developed. The authors optimized the projector configuration. A system that used 300 projectors and a 100-in. screen was developed, with a projection length of 3.4 m. To obtain a large screen size, the multi-projection system requires a long projection length; thus, a large operating space is required. The complexity of this setup makes installation time-consuming and relocation difficult. Moreover, a dark-room environment is required. The tiling of small-screen 3D displays has also been proposed for the construction of large-screen, glasses-free 3D displays. Tiling systems do not require a long projection length because each display has a small screen. In Refs. 8 and 9, images projected by small projectors are tiled on a large parallax barrier. In Ref. 10, a 3D pixel module was constructed using a two-dimensional (2D) array of liquid crystal display (LCD) panels and a 2D array of cylindrical lenses; these 3D pixel modules were then two dimensionally tiled. In Ref. 11, tiled multi-view display modules were used, which were composed of MEMS projectors and a lenticular lens. The above tiling techniques eliminated gaps existing in tiled screens. Reference 12 presented a technique of tiling conventional multi-view displays. Because of the bezel of each multi-view display, the tiled screen had apparent vertical and horizontal gaps.
Fig. 1. Tiled large-screen 3D displays: (a) landscape, (b) portrait, and (c) curved.
In this paper, we propose the tiling of frameless multi-view display modules that employ flat-panel displays. Compared with the multi-projection system, the proposed system requires less space. At present, high-resolution flat-panel displays are being available at more affordable costs. The proposed system does not require a dark room, and the tiled screen can be configured in various ways such as landscape, portrait, or curved, as shown in Fig. 1. Therefore, it can be used in various applications. 2. Frameless multi-view display module In this study, a multi-view flat-panel display, which consists of a flat-panel display and a lenticular lens, is used for 3D image generation. When these multi-view flat-panel displays are tiled, the display bezels generate horizontal and vertical black bars in the tiled screen. In this study, the multi-view flat-panel display is combined with an imaging system in order to obtain a frameless screen.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6211
As shown in Fig. 2, the proposed multi-view display module consists of a multi-view flatpanel display, an aperture, an imaging lens, a screen lens, and a vertical diffuser. Two imaging systems exist in the module: one is used for projecting the screen and the other for the viewpoints. In the first imaging system, the imaging lens projects the screen of the multi-view flatpanel display onto that of the module, the latter of which consists of a screen lens and a vertical diffuser. When the magnification is larger than unity, a frameless screen is obtained. The imaging system is designed such that the dimensions of the imaging lens do not exceed those of the screen of the module.
Fig. 2. Proposed multi-view display module with frameless screen.
Fig. 3. Operating principle of the proposed frameless multi-view module: (a) horizontal sectional view, and (b) vertical sectional view.
The second imaging system is explained as follows. The aperture is placed on the plane where the viewpoints of the multi-view flat-panel display are generated. The combination of the imaging and screen lenses projects the viewpoints of the multi-view flat-panel display onto the observation space in order to generate viewpoints for the observers, as shown in Fig. 3(a). The aperture eliminates repeated viewpoints in the horizontal direction. Therefore, the 3D images do not flip at the boundary of the viewing area. The aperture also limits the rays in the vertical direction, which helps reduce image degradation due to aberrations, as shown in Fig. 3(b). When the viewpoints of the multi-view flat-panel display are generated at the opposite side across the imaging lens, the position and size of the aperture are determined by considering the image of the viewpoints formed by the imaging lens. As aperture height decreases, the height of the viewing area decreases. The vertical diffuser attached to the screen lens is used to increase the vertical viewing area. When multiple modules are tiled to obtain a large screen size, the viewing areas of all modules must coincide to create a common viewing area. Three possible methods are available. The first is to properly shift the screen lens in each module in order to produce a common viewing area, as shown in Fig. 4(a). The second is to properly shift the aperture in each module, as shown in Fig. 4(b). The third method, as shown in Fig. 4(c), involves properly rotating each module. The aforementioned three methods can be used simultaneously. #205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6212
Fig. 4. Generation of a common viewing area for tiled modules: (a) shifting of the screen lens, (b) shifting of the aperture, and (c) rotation of the modules.
3. Prototype system The frameless multi-view display module was designed, and four modules were tiled vertically to obtain a large screen that can display a human-sized object. This study employed an LCD panel with a resolution of 3,840 × 2,400 and a screen size of 22.2 in. Because large dimensions are required for the imaging and screen lenses, Fresnel lenses were used. In general, Fresnel lenses do not possess ideal imaging properties. We experimentally designed the imaging system for the screen imaging by using commercial Fresnel lenses and evaluated the distortion and blur of the images. Consequently, we decided to construct the imaging system by using two Fresnel lenses with a measured focal length of 592 mm. The designed imaging system is shown in Fig. 5. The screen size of the module was 27.3 in., and the magnification was 1.23. The length of the imaging system was 1.49 m. A slanted lenticular lens [9] was designed, which was attached to the LCD panel to construct the multi-view flat-panel display. The multi-view flat-panel display was designed with a resolution of 320 × 200 and 144 viewpoints. The slant angle of the lenticular lens was tan−1(1/12) = 4.76°. The viewpoints were generated with an interval of 2.14 mm at a distance of 537 mm from the surface of the multi-view display. As shown in Fig. 5, the viewpoints were imaged by the left imaging Fresnel lens to generate viewpoints with an interval of 2.02 mm at a distance of 535 mm from the surface of the multi-view display. At this position, a 291-mm-wide aperture was placed to eliminate repeated viewpoints; this aperture had the same slant angle as that of the lenticular lens. The aperture height (95.0 mm) was determined experimentally with a view to reduce image distortion and blur caused by aberrations. A commercially available Fresnel lens with a measured focal length of 991 mm was used as the screen lens. The combination of the screen and imaging Fresnel lenses generates viewpoints for viewers with an interval of 18.3 mm at a distance of 5.79 m from the screen. The horizontal width of the viewing area at this distance was 2.64 m. To realize a frameless screen, the side surfaces of the screen Fresnel lens were processed to have two-step structures,
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6213
as shown in Fig. 6. Grooves of depth 1.0 mm were used to support the screen Fresnel lens by using the side plates of the module to obtain the module’s required flat side surfaces.
Fig. 5. Design of the frameless multi-view display module: (a) horizontal sectional view and (b) vertical sectional view.
Fig. 6. Screen Fresnel lens having side surfaces with two-step structures.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6214
Fig. 7. Constructed frameless multi-view display module.
A lenticular lens having a pitch of 1.48 mm and a slant angle of 4.76° was used as the vertical diffuser, in order to increase the vertical viewing area. Without the vertical diffuser, the height of the viewing area was 0.814 m at a distance of 5.79 m from the screen. Figure 7 shows the constructed multi-view display module (the side plates have been removed to show the inside of the module). The screen size of the multi-view module was 589 mm × 368 mm (27.3 in.), and the depth of the module was 1.5 m. The module had a frameless screen and flat side surfaces. A prototype display system with a medium-sized screen was constructed using four modules, which were vertically aligned to obtain a screen size of 62.4 in. (589 mm × 1,472 mm). To generate a common viewing area for all modules, the centers of the screen lenses were shifted in the vertical direction by −109.2, −36.4, + 36.4, and + 109.2 mm for the modules located from the top to the bottom. Figure 8 shows the tiled display system where human-sized 3D images are displayed.
Fig. 8. Vertically aligned four modules.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6215
4. Distortion correction Because Fresnel lenses were used to construct the multi-view display modules, distortions were present in both the screen imaging and viewpoint generation. The present study electronically corrected these distortions. The distortion in the screen imaging is shown in Figs. 9(a)–9(c). Identical checkerboard patterns were displayed to all viewpoints, and the images were captured at the leftmost, center, and rightmost positions in the viewing area at the distance where the viewpoints were generated. As shown in Fig. 9(b), obvious distortion was not present at the center position. On the other hand, the vertical-line interval changes in the horizontal direction at the left- and rightmost positions, as shown in Figs. 9(a) and 9(c), respectively. Figures 10(a)–10(c) show the distortion in the viewpoint generation. The checkerboard pattern was displayed only at the center viewpoint. Figure 10(b) shows the image captured at the center position in the viewing area, and only the center part of the pattern was observed. Figures 10(a) and 10(c) show the images captured at the left- and right-side positions, respectively. If rays were correctly converged to the center viewpoint, the entire checkerboard pattern would be observed at the center viewpoint. The results showed that rays from the entire screen did not correctly converge to the center viewpoint, and rays from the left and right parts of the screen went to the left and right viewpoints. Therefore, the viewpoint generation contains distortion. To correct the distortions, they were measured and then described using approximate equations. We considered a ray that should be emitted from the position (x, y) on the screen to a viewpoint in which the horizontal position is u on the plane where the viewpoints were generated. From the measurement, we found that an actual ray was emitted from the position (X, Y) on the screen and approached a horizontal position U. Multiple correspondences between (x, y, u) and (X, Y, U) were measured. Three functions x = f (X, Y, U), y = g (X, Y, U), and u = h (X, Y, U) were estimated from the measured correspondences. Functions f, g, and h were represented by polynomial equations, and the coefficients were determined by the least squares method. The number of measured data points is denoted by n. The measured data are represented by (Xi, Yi, Ui) corresponding to the display data (xi, yi, ui), where 0 ≤ i ≤ n − 1. In this study, the error function was defined as follows: n −1
E= i =0
{ x − f ( X ,Y ,U ) + y − g ( X ,Y ,U ) + u − h ( X ,Y ,U ) }. 2
i
i
i
i
2
i
i
i
i
2
i
i
i
i
(1)
The three functions are represented by m-th order polynomial functions. f ( X , Y ,U ) = g ( X , Y ,U ) =
h ( X , Y ,U ) =
i + j + k ≤m
i , j ,k ≥0 i + j +k ≤m
i , j ,k ≥0
i + j +k ≤m
i , j ,k ≥ 0
ai , j ,k X iY jU k ,
(2)
bi , j ,k X iY jU k ,
(3)
ci , j ,k X iY jU k .
(4)
The coefficients ai, j, k, bi, j, k, and ci, j, k were determined by minimizing the error function using the least squares method. The Gauss–Jordan elimination was used to solve the problem.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6216
Fig. 9. Correction of the distortion in the screen imaging: (a)–(c) without correction and with correction using (d)–(f) quadratic approximation and (g)–(i) cubic approximation. (a), (d), and (g) were captured from the leftmost viewing position; (b), (e), and (h) were captured from the center viewing position; and (c), (f), and (i) were captured from the rightmost viewing position.
Fig. 10. Correction of the distortion in the viewpoint formation: (a)–(c) without correction and with correction using (d)–(f) quadratic approximation and (g)–(i) cubic approximation. (a), (d), and (g) were captured from the left-side viewing position; (b), (e), and (h) were captured from the center viewing position; and (c), (f), and (i) were captured from the right-side viewing position.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6217
To find the correspondences between (x, y, u) and (X, Y, U), a grid pattern was displayed on a certain viewpoint, and several intersection points of the projected grid pattern were found from several viewing positions in the viewing area by using a digital camera. The camera was placed at horizontal positions U = −800, −400, 0, + 400, and + 800 mm on the plane where the viewpoints were generated, and the grid pattern was displayed at viewpoints #0, #5, #10, …, #140. Several intersection points (X, Y) were measured for each captured image. Approximately 400 correspondences were obtained. Functions f, g, and h were approximated using quadratic or cubic equations, which had 10 or 20 coefficients. The number of correspondences was large enough to determine the coefficients by using the least squares method. Once the approximate equations were determined, the images displayed by the multi-view flat-panel display were synthesized using these equations. The images (x, y) displayed on horizontal position u were interpolated from the parallax images (X, Y) corresponding to the viewpoints at horizontal position U. Figures 9(d)–9(f) show the corrected images captured at the three positions when the checkerboard patterns were displayed to all viewpoints. The functions were approximated using quadratic equations (m = 2). Figures 9(g)–9(i) show those approximated using cubic equations (m = 3). The horizontal variation in the intervals of the vertical lines decreased at the leftmost and rightmost viewing positions. To show the effects of the correction, intervals of the vertical lines were measured at the leftmost and rightmost viewing positions. The intervals among the central five lines were measured. Before the correction, from Figs. 9(a) and 9(c), the minimum and maximum intervals were 39.0 and mm 34.4 mm, respectively. When quadratic approximation was used, from Figs. 9(d) and 9(f), the intervals were 38.4 and 36.4 mm. When cubic approximation was used, from Figs. 9(g) and 9(i), they were 38.4 and 36.4 mm. Figures 10(d)–10(f) and 10(g)–10(i) show the corrected images captured at the three viewing positions when the image was displayed only at the center viewpoint, and the functions were approximated using quadratic and cubic equations, respectively. A more checkerboard pattern was observed at the center viewpoint, and no images were observed at the left and right viewpoints. The results showed that the distortion of the viewpoint generation was reduced. We could find little difference between corrected images generated using quadratic and those using cubic approximations. 5. Three-dimensional image generation by four modules
The 3D images were displayed using the tiled 3D display consisting of four vertically aligned modules. The distortion correction described in the previous section was performed for all modules. The camera used for the distortion correction was positioned in the common viewing area, and cubic approximation was used. After the correction, the width of the viewing area where an entire 3D image could be observed was approximately 2.3 m at a distance of 5.79 m from the screen. When the viewing distance became larger than 5.79 m, the viewing width increased. When the viewing distance became smaller, the viewing width decreased. Figure 11 shows the 3D images generated by the four modules. Figures 11(a)–11(c) show the images without the distortion correction, and Figs. 11(d)–11(f) show those with correction. With the correction, the connectivity of the images among the modules was improved. However, the connection of the images between the top and second modules worsened when viewed from the left position. Without the correction, the shape of the space shuttle appeared rounded inward; the left and right sides were further observed. With the correction, the image blur was decreased, and the vertical stabilizer appeared thinner. Figure 12 shows the other 3D images. Smooth motion parallax was obtained, which can be observed in the provided movies.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6218
Fig. 11. 3D image generated by the vertically aligned four modules: (a)–(c) without correction, and (d)–(f) with correction using cubic approximation. (a) and (d) were captured from the left viewing position, (b) and (e) were captured from the center viewing position, and (c) and (f) were captured from the right viewing position.
6. Discussion
In the photographs shown in the previous section, thin gaps were observed at the boundaries between the modules. The imaging system of the module was experimentally designed such that it provided an image larger than a screen of the module. We checked that the rays were not vignetted by the side panels of the module using paraxial approximation. However, because of the field curvature of the imaging system, rays came into focus before the screen lens. Thus, some rays were vignetted by the side panels and side surfaces of the screen lens. Therefore, the light intensity near the screen boundaries decreased. The field curvature can be reduced when aspheric surfaces are used as structures of the Fresnel lenses. While using commercial Fresnel lenses, it is difficult to completely eliminate the gaps between modules. The Fresnel lenses with aspheric surfaces have to be designed for multi-view display modules to reduce field curvature. The vertical diffuser was attached to the screen lens to increase the vertical viewing area. Without the vertical diffuser, the entire 3D image was observed at distances near the plane where the common viewpoints were generated, i.e., at 5.79 m from the screen, because the height of the viewing area was 0.814 m at this distance. However, at distances far from this plane, partial 3D images were observed, and the upper and lower parts of the tiled screen appeared black. By adding the vertical diffuser, the entire 3D image could be observed at any distance. However, the brightness of the 3D images decreased with the vertical diffuser. Figure 13 shows 3D images when the vertical diffusers were removed from display modules. The sharpness of the 3D images decreased because the vertical diffuser caused vertical blurs in the 3D images.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6219
Fig. 12. 3D images generated by the vertically aligned four modules: (a)–(c) pilot (Media 1), (d)–(f) girl (Media 2), and (g)–(i) space shuttle (Media 3). (a), (d), and (g) were captured from the left viewing position; (b), (e), and (h) were captured from the center viewing position; and (c), (f), and (i) were captured from the right viewing position.
Fig. 13. 3D images generated by the vertically aligned four modules without vertical diffusers: captured from (a) left-, (b) center-, and (c) right-viewing positions.
The 3D images generated by the tiled display were bright enough to be viewed in a room at daytime, even with the vertical diffuser attached to the screen. The vertical diffuser caused diffused reflection of environmental light, reducing the contrast in the 3D images. The vertical diffuser degraded 3D images as mentioned above. Instead of using the vertical diffuser, the vertical viewing area can be increased by increasing the height of the aperture in the imaging system. However, when the aperture height increases, the aberrations of the imaging system increase, resulting in blurs and distortions of 3D images. Therefore, the Fresnel lenses should be designed to reduce the aberrations of the imaging system in order to allow the increase in aperture height. Correction of the distortion in the screen imaging has been applied for multi-projection systems [13], and correction of the viewpoint generation has also been applied for multiprojection systems [14]. In the tiled display system proposed in this paper, the former #205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6220
distortion type causes image discontinuity between adjacent modules, and the latter causes shape distortion of 3D objects and blurs in 3D images. In our prototype system, the former distortion type was not large as shown in Fig. 9, whereas the latter was large as shown in Fig. 10. The effects of the distortion correction are shown in Fig. 11. Because the distortion was not completely removed, more precise correction is required by increasing the degree of polynomials and the number of correspondences. The display depth range of 3D images is determined by the screen size. In our system, it was also determined by the blurs in the 3D images. The blurs were caused by the distortion in the viewpoint generation and the crosstalk among viewpoints. Without the distortion correction, the depth range with allowable blurs was within a range of ± 600 mm. With the distortion correction, the depth range was increased to within a ± 1,000-mm range. The depth of the tiled 3D display system is determined by the length of the imaging system of the modules. The length of the imaging system could be reduced by using more Fresnel lenses. The use of more Fresnel lenses would reduce the aberrations. However, the use of too many Fresnel lenses does not effectively reduce the length of the imaging system. 7. Conclusion
In this study, the tiling of multi-view display modules was proposed to construct large-screen glasses-free 3D displays. The module was designed to have a frameless screen, which consisted of a multi-view flat-panel display and a projection imaging system. The constructed modules had a screen size of 27.3 in. and a resolution of 320 × 200. Four modules were aligned vertically to provide a screen size of 62.4 in. for displaying a human-sized object. The distortion correction in the screen imaging improved the connectivity of images between adjacent modules. Furthermore, the distortion correction of the viewpoint generation improved the shape reproduction of the 3D objects and reduced blurs in 3D images. Acknowledgment
The 3D data of “pilot” shown in Figs. 8 and 11 was created by Hero Nguav and edited by Howard Day, and that of “space shuttle” shown in Figs. 10 and 11 was distributed by NASA.
#205530 - $15.00 USD (C) 2014 OSA
Received 30 Jan 2014; revised 27 Feb 2014; accepted 3 Mar 2014; published 10 Mar 2014 24 March 2014 | Vol. 22, No. 6 | DOI:10.1364/OE.22.006210 | OPTICS EXPRESS 6221
