In this paper, we propose new auto-stereoscopic 3D displays that substantially reduce crosstalk. In general, it is difficult to eliminate crosstalk in auto-stereoscopic 3D displays. Ideally, the parallax barrier can eliminate crosstalk for a single viewer at the ideal position. However, due to variations in the viewing distance and the interpupillary distance, crosstalk is a problem in parallax barrier displays. In this paper, we propose 3-dimensional barriers, which can significantly reduce crosstalk.
©2011 Optical Society of America
Several auto-stereoscopic 3D display techniques have been proposed to create 3D images [1,2]. Typical auto-stereoscopic displays use a screen (parallax barrier) or lens (lenticular) system to register correct images to each eye [3–7].
After a single-panel 3D display based on the parallax barrier was proposed , substantial research has been conducted to develop better auto-stereoscopic 3D displays [9–15]. By using various types of parallax barriers, several single-viewer stereoscopic displays have been developed [8,10]. Although conventional auto-stereoscopic displays using parallax barriers allow viewers to perceive 3D images without requiring any special glasses, the crosstalk due to variations in the viewing distance and viewing angle may cause eye fatigue and perceptual quality degradation. On the other hand, to reduce the crosstalk due to the movements of viewers, 3D displays with adjustable barriers have been proposed [14,16]. As the viewing distance changes or the viewer moves, the display tracks eye movements and improves the viewing distance range by adjusting the barriers. However, adjusting the barriers requires a tracking sensor and tracking errors may occur, which may result in inaccurate adjustments. Recently, Yoon et al. proposed to use the Lucius prism, which is a triangular prism having a reflective or black film on one side for auto-stereoscopic 3D display . They showed how the Lucius prism can be fabricated and directionally allocate the incident light intensity. However, the identically shaped prism array may generate crosstalk.
Lenticular optical systems include a transparent sheet and narrow vertical lenses, which produce a three-dimensional effect . Although these systems can provide a multi-view image with higher brightness than parallax barrier systems, there are some drawbacks. For example, viewers must sit in specific spots and incorrect lens alignments may cause undesirable distortions [19,20]. Recently, a stereoscopic 3D LCD device was proposed . This device uses a double sided prism film and a scanning backlight. Although the scanning backlight improves performance, it is still difficult to completely eliminate the crosstalk problem with this kind of film-type 3D display. Also, this method requires a scanning backlight and therefore may not be applicable to display devices that do not have these backlight functions such as OLED (Organic Light-Emitting Diode).
In this paper, we propose new 3-dimensional barriers, which effectively reduce the amount of crosstalk in auto-stereoscopic 3D displays. Assuming that sub-pixels are alternatively placed for the left and right eyes in the horizontal direction, we constructed V-shaped barriers. The width, height and angle of the V-shaped barriers were automatically determined for a given display with a fixed pixel width, the viewing distance and the interpupillary distance (IPD) of the viewer. Simulations show that the proposed V-shaped barriers can reduce the amount of crosstalk almost by half.
2. Autostereoscopic displays with 3-dimensional barriers
2.1 Constructing 3-dimensional barriers for autostereoscopic displays
To perceive 3D objects, the two eyes must see different images. One possible way to achieve this effect is to place sub-pixels alternatively for the left and right eyes in the horizontal direction and control the outgoing light from the sub-pixels. In Fig. 1 , the light from the sub-pixels for the left eye (white) should reach only the left eye while the light from the sub-pixels for the right eye (gray) should reach only the right eye. In traditional parallax barrier displays, a number of vertical stripe barriers are placed in front of the display as shown in Fig. 1.
In the proposed 3D display, we place V-shaped barriers as shown in Fig. 2 . It can be seen that these V-shaped barriers block the light from the sub-pixels for the left eye from reaching the right eye and the light from the sub-pixels for the right eye from reaching the left eye. Figure 3 shows a 3D view of the V-shaped barriers.
The width, height and angles of the V-shaped barriers are determined depending on the pixel width, the viewing distance and the interpupillary distance. In particular, the angles of the V-shaped barriers vary depending on the horizontal location of the pixels as illustrated in Fig. 4 .
We can easily compute the angles of the V-shaped barriers. We set the center of the display as the origin (Fig. 4). Assuming that the viewing distance is , the interpupillary distance , and the pixel width of the display , the line that passes the left eye at and is given by
The line that passes the right eye at and is given by
Then, the n-th inersection can be computed by
Using Eq. (3), one can easily compute the angles of the V-shaped barriers. Equation (3) also shows that the width and height of the barriers are automatically determined once the viewing distance, the pixel width, and the interpupillary distance (IPD) are given. Various manufacturers produce their 3D displays for the most representative user parameters (interpupillary distance and viewing distance). It can be seen that the height (the value of y) of the V-shaped barrier (and the parallax barrier) is independent of n. It is a function of the viewing distance, the pixel width, and the interpupillary distance. In general, the pixel width and barrier height are fixed for 3D displays. Also, the interpupillary distance cannot be adjusted. The only adjustment that users can make is the viewing distance. Since x in Eq. (3) is a function of and , different interpupillary distances or slight movements from the ideal position will always cause crosstalk. If a 3-D display is designed for a single user and if the user is located at the ideal position, there will be no difference between the proposed 3D display with V-shaped barriers and the conventional parallax barrier 3D display. However, if the eye positions change due to personal variations in the interpupillary distance or if the viewing position changes, crosstalk always occurs in both cases, although the 3D display with adjustable barriers [14,16] may compensate for the viewing distance change and personal variations in the interpupillary distance to some degree.
2.2 Crosstalk comparison
In parallax barrier 3D displays, the amount of pixel crosstalk can be measured as follows:
Then, the crosstalk of a 3D display can be defined as an average of the entire set of pixels. The proposed 3D display substantially reduces the crosstalk problem compared to conventional parallax barrier 3D displays. Figure 5 illustrates how this reduction is achieved. As the right eye moves to the right, it sees portions of the pixels intended for the left eye. However, the V-shaped barriers effectively block this crosstalk, as can be seen in Fig. 5. Also, if the right eye moves to the left, the conventional parallax barrier and the V-shaped barrier generate the same amount of crosstalk. Although the V-shaped barriers cannot completely eliminate the crosstalk, they can reduce it almost by half.
Figure 6 . (a) shows the crosstalk ratio of a conventional 3D display with parallax barriers as the eye moves away horizontally and toward/away from the ideal position. It was assumed that the interpupillary distance was 65mm, the viewing distance was 40cm, the pixel size was 0.217mm, and the number of pixels per line was 2,000 (e.g., typical PC monitors, total display width = 43.4cm). The display height does not affect the crosstalk ratio. Figure 6. (b) shows the crosstalk ratio of the proposed 3D display. It can be seen that the proposed V-shaped barrier substantially reduced the crosstalk. Similar improvements were observed at other viewing distances (e.g., 50cm, 60cm, etc.). Tables 1(a) and 2(a) show the amount of crosstalk improvement of the proposed 3D display with V-shaped barriers compared to a conventional 3D display with parallax barriers. In these cases, the target IPD of the displays are 65mm (i.e, the displays are designed for viewers whose IPDs are 65mm) and it is assumed that the viewer’s IPD is also 65mm. This improvement was measured by dividing the amount of crosstalk of the proposed display by the amount of crosstalk of the conventional display. It can be seen that the amount of crosstalk was reduced by about half.
On the other hand, there would be no crosstalk reduction for a viewer with a smaller IPD assuming that the viewer watches at the ideal position (Table 1(b), Table 2(b)). However, as the viewer moves slightly from the ideal position, the viewer also would experience crosstalk reduction. A larger crosstalk deduction would be achieved for a viewer with a large IPD (Table 1(c), Table 2(c)).
Recently, some portable terminals have also been built to support 3D displays. One characteristic of these 3D displays is that the display width can be smaller than or roughly equal to the interpupillary distance. Figure 7 shows the amount of crosstalk of a portable 3D display (pixel width: 0.077 mm, display width: 7.7cm (the number of pixels per line: 1,000), viewing distance: 40cm). Similar improvements were observed at other viewing distances (Table 3 ).
In summary, the amount of crosstalk of the proposed display is about 49.73% of the conventional parallax barrier display for 3D monitors (i.e., the display width is larger than the interpupillary distance) and 49.66% for mobile terminals as the eyes move in the toward/away direction to the display [35 cm, 45 cm] and the horizontal transition [-5 cm, 5 cm]).
2.3 Illumination loss
As the eyes move away from the ideal position, some illumination loss also occurs (Fig. 5). In other words, the image becomes darker when using the proposed display. This illumination loss is identical for both the conventional parallax barrier and the proposed 3-dimensional barrier within the intended viewing area (Fig. 8 ) of the proposed method. However, when the eyes move further away from the ideal position, the proposed auto-stereoscopic display with 3-dimensional barriers looks completely dark. For quantitatively analyses, we define the left-eye transmittance as the amount of light from the pixels for the left eye that actually reaches the left eye (the sub-area of a pixel which the viewer sees). Similarly we can define the right-eye transmittance. Ideally, these transmittances should be 1 or 100%.
Figure 9 shows the transmittance and crosstalk for the left eye. As can be seen, if the left eye moves more than 13cm either to the left or to the right (Figs. 9(a,c)), the display with the proposed 3-dimensional barriers looks completely dark. On the other hand, the conventional display with parallax barriers has several viewing spots (Fig. 9(b)). However, it still suffers from the same crosstalk problem. In other words, with a slight move from the ideal position, the amount of crosstalk rapidly increases. The proposed method successfully reduced this crosstalk by half within the intended viewing area. If the 3D display is intended for a single use, this characteristic of the proposed display is not a serious drawback. In particular, if the ratio of the crosstalk to the transmittances is large, 3D effects will rapidly decrease. For instance, if the crosstalk-transmittance ratio is greater than 1, there would be no 3D effects at all (Figs. 9(e~h)). In Figs. 9(e~h), if the ratio is larger than 1, it is clipped to 1. For the right eye, similar results were obtained (Fig. 10 ). The total crosstalk is the sum of the left and right eye crosstalk (Fig. 11 ). In conventional 3D displays, a small horizontal move (about 3.3cm either to the right or to the left) makes the crosstalk-transmittance ratio larger than 1, resulting in no 3D effects (Fig. 11) while the proposed method reduces the ratio by half. For example, with 1cm shift, the crosstalk-transmittance ratio of the conventional display is 18% while it is 9% for the proposed display.
Also viewers may move forward or backward from the display. In this case, the transmittance and crosstalk of the left and right eyes will be identical due to the symmetry. Figure 12 shows a comparison of transmittances, crosstalk, and crosstalk-transmittance ratios for one eye. Figure 13 shows a comparison of the total crosstalk-transmittance ratios within the intended viewing area. Again, the proposed method significantly reduced the amount of crosstalk.
2.3 Illumination loss
In practice, it is rather difficult to exactly manufacture the V-shaped barriers. However, we can obtain the identical benefit of reducing the amount of crosstalk by placing any vertical barriers at the pixel boundaries of the left and right sub-pixels (Fig. 14 ). Most of these vertical barriers do not meet the centers of the horizontal barriers if the vertical barriers are perpendicular to the display. Also, at both ends, the vertical barriers may not be made to be perpendicular to the display. Furthermore, the vertical barriers need not reach the horizontal barriers. If they are adequately tall, they are still able to remove the crosstalk as long as the eye location changes are not too large (as shown by the partial vertical barrier in Fig. 14). The effective height of the vertical barriers depends on the size of the target viewing regions.
On the other hand, typical display panels have protective glass that covers light emitting devices. Thus, if the proposed vertical barriers are placed on the glass, some crosstalk may occur due to the leak through the glass. However, as display technologies advance, this problem can be reduced or eliminated. For example, manufacturers are exploring the possibility of using plastic or flexible panels instead of glass panels for displays. For these kinds of panels, it would be possible to etch vertical walls so that they actually reach the light emitting unit boundaries.
In this paper, we have proposed new auto-stereoscopic displays with 3-dimensional barriers. Compared to the existing parallax barrier 3D display, the proposed auto-stereoscopic displays can significantly reduce the amount of crosstalk without causing any additional illumination loss. We presented some implementation issues of the proposed 3D displays with some computer simulation results. Using the proposed 3D displays, it is possible to build auto-stereoscopic 3D displays which can reduce the amount of crosstalk almost by half when there are variations in viewing distances and interpupillary distances.
This work was supported in part by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2011-0029381).
References and links
1. T. Okoshi, Three-Dimensional Imaging Techniques (Academic, 1976).
2. T. Okoshi, “Three-dimensional displays,” Proc. IEEE 68(5), 548–564 (1980). [CrossRef]
3. N. A. Dodgson, “Autostereoscopic 3D displays,” Computer 38(8), 31–36 (2005). [CrossRef]
4. I. Sexton and P. Surman, “Stereoscopic and autostereoscopic display systems,” IEEE Signal Process. Mag. 16(3), 85–99 (1999). [CrossRef]
5. L. Hill and A. Jacobs, “3-D liquid crystal displays and their applications,” Proc. IEEE 94(3), 575–590 (2006). [CrossRef]
6. G. Lawton, “3D displays without glasses: coming to a screen near you,” Computer 44(1), 17–19 (2011). [CrossRef]
7. H. Urey, K. V. Chellappan, E. Erden, and P. Surman, “State of the art in stereoscopic and autostereoscopic displays,” Proc. IEEE 99(4), 540–555 (2011). [CrossRef]
8. D. Ezra, G. J. Woodgate, B. A. Omar, N. S. Holliman, J. Harrold, and L. S. Shapiro, “New autostereoscopic display system,” Proc. SPIE 2409, 31–40 (1995). [CrossRef]
9. Y. Kajiki, H. Yoshikawa, and T. Honda, “Autostereoscopic 3-D video display using multiple light beams with scanning,” IEEE Trans. Circ. Syst. Video Tech. 10(2), 254–260 (2000). [CrossRef]
11. C. M. G. Lee, A. R. L. Travis, and R. Lin, “Flat-panel autostereoscopic 3D display,” IET Optoelectronics 2(1), 24–28 (2008). [CrossRef]
12. G. J. Woodgate, D. Ezra, J. Harrold, N. S. Holliman, G. R. Jones, and R. R. Moseley, “Autostereoscopic 3D display systems with observer tracking,” Signal Process. Image Commun. 14(1–2), 131–145 (1998). [CrossRef]
13. H. Yamamoto, M. Kouno, S. Muguruma, Y. Hayasaki, Y. Nagai, Y. Shimizu, and N. Nishida, “Enlargement of viewing area of stereoscopic full-color LED display by use of a parallax barrier,” Appl. Opt. 41(32), 6907–6919 (2002). [CrossRef] [PubMed]
14. T. Peterka, R. L. Kooima, D. J. Sandin, A. Johnson, J. Leigh, and T. A. DeFanti, “Advances in the Dynallax solid-state dynamic parallax barrier autostereoscopic visualization display system,” IEEE Trans. Vis. Comput. Graph. 14(3), 487–499 (2008). [CrossRef] [PubMed]
16. S.-Y. Wu and W.-Y. Chen, “Adjustable parallax barrier 3D display,” U.S. Patent 20100060983 (2010).
17. H. Yoon, S.-G. Oh, D. S. Kang, J. M. Park, S. J. Choi, K. Y. Suh, K. Char, and H. H. Lee, “Arrays of Lucius microprisms for directional allocation of light and autostereoscopic three-dimensional displays,” Nat Commun. 2, 455 (2011). [CrossRef] [PubMed]
18. M. G. Lippmann, “Epreuves reversibles donnant la sensation du relief,” J. Phys. 7(4), 821–825 (1908).
19. B. A. Rosenthal, “Lenticular optical system,” U.S. Patent 7,724,438 (May 25, 2010).
20. Y. Abe, “Lenticular lens sheet and production method therefor,” U.S. Patent 7,477,450 (Jan. 13, 2009).
21. S. E. Brigham, J. C. Schultz, and B. L. Weaver, “Stereoscopic 3D liquid crystal display apparatus with scanning backlight,” U.S. Patent application 2008/0084519 A1 (Apr. 10, 2008).