Wearing special glasses to watch 3D video is inconvenient for many people and especially those who wear glasses for the correction of eyesight. Moreover, wearing glasses is not natural for applications such as role-playing games, virtual reality, and mobile displays. Therefore, high-quality ‘autostereoscopic’ displays (i.e., that do not require glasses) are desirable. An autostereoscopic display projects different images of the same object onto the left and right eyes through independent spatial channels. Conventional autostereoscopic displays include ‘lenticular’ and ‘parallax barrier’ (PB) displays. Lenticular 3D displays use a cylindrical lenticular lens to direct diffuse light from an individual pixel so it can only be seen in a limited angle in front of the display. This then allows different images to be directed to the viewer's left or right eye. Existing PB display designs use a barrier in front of a screen such as an LCD display. Slits in the barrier separate left and right images displayed interlaced on the screen, and send them into left and right eyes of the viewer, respectively (see Figure 1). However, the inherent structure of conventional autostereoscopic displays, such as PB or lenticular 3D displays, requires the viewer to remain in a position such that each eye is in the appropriate viewing zone; if they do not, ‘crosstalk’ between adjacent viewing zones mean that each eye will see both left and right images simultaneously (see the example of the inset pair of eyes in Figure 1).
To overcome this, researchers have previously proposed a ‘dynamic’ (moving) PB with segmented slits,1 or a dual-stacked LCD display.2 A dynamic PB with segmented slits divides a conventional parallax barrier slit into several segments, where electrical switches control the tiny moving parallax barrier (see Figure 2). However, light leaks between slit segments and degrades the image. A dual-stacked LCD display is constructed by stacking multiple layers of liquid crystal, and its displayed images are changed according to the viewer's position. It is necessary to improve the low brightness, low effective resolution, and barrier line visibility with this design.
Figure 1. Viewing zones for a conventional single-layer parallax barrier. When viewers have their left eyes in a red zone and their right eyes wholly in a blue zone, they will see a good 3D image. If, however, their eyes are in the position of the inset pair of eyes, each eye will simultaneously see the left and right images.
Recently, we demonstrated a novel advanced autostereoscopic 3D display, which we have named the dual-layer parallax barrier (DLPB). It can shift the position of the viewing zones without heavily increasing the viewing distance by combining a dual PB, dual pairs of indium tin oxide (ITO), and a liquid crystal (LC). In addition to providing dynamic viewing zones, the DLPB's independent ITO pairs sandwich a common LC, which reduces its thickness. Moreover, when connected to a head tracking system, it provides a smooth transition of 3D depth feeling as the viewer changes position (Figure 2).
Figure 2. White light pattern images operated by the first and the second parallax barrier (PB) layers (a) the first PB layer (b) the second PB layer. ITO: Indium tin oxide. TN-LC: Twisted nematic liquid crystal.
We used the commercially available thin-film transistor liquid crystal display (TFT-LCD) LMS480KF01 made by Samsung. It is 4.8" across, and has a pixel pitch of 0.12975mm and a resolution of 800×480. The block width and the slit width of each PB layer are 0.17262mm and 0.08632mm, respectively. The distances between each PB layer and TFT-LCD are 0.863mm and 0.007mm. Finally, the offset displacement between the two PB layers is 0.06474mm. Compared to conventional PB displays, the increase in thickness of the DLPB display from adding two ITO layers and two insulator layers is only 0.007mm, which is reasonable for commercialization.
Figure 3. White light pattern images are shifted horizontally when the second PB is used, compared to when the first is used.
Figure 4. Illustration of multi-view 3D video service complying with each viewing angle according to the movement of viewers.
The DLPB method is able to change the position of the aperture in a twisted nematic LC (TN-LC) display by controlling the polarization of light. Our method applies a voltage to one of the pairs of ITO. This sets up a local electric field in the TN-LC and, as a result, twists the polarization of vertically polarized light from a TFT-LCD panel so that it is horizontally polarized and can pass through a horizontal polarizer (see Figure 2). This determines which path is taken by light from the display and shifts the light emerging from the DLPB such that the positions of the viewing zones are moved when the second (rather than the first) PB is switched on (see Figure 3).
Figure 5. Block diagram of the prototype of the dual-layer parallax barrier (DLPB) system. The viewing zone numbers refer to four cases for the interlacing order and the choice of PB: #1 is PB1 and left image first; #2 is PB2 and left image first; #3 is PB1 and right layer first; and #4 is PB2 and right layer first, for a seven-position setup.
We have also designed a DLPB system prototype that includes a video camera and a laptop-based head-tracker. The video camera captures images that are analyzed by the laptop to calculate the viewer's position, and then it optimizes the viewing zones for that location in real time. This is done by selecting the order in which to interlace the images shown on the display (left first or right first) and which PB to turn on. In addition to optimizing the viewing zones, it is possible to select the images so that the angle of view matches that of the viewer (see Figure 4). Our prototype has a total of seven viewing positions, 44 images viewed from separate angles with 800×480 resolution, and a horizontal field of view of 29.394° (see Figure 5). The threshold values of the boundaries between viewing zones (according to the different beam forming states) are −14.697°, −10.611°, −6.413°, −2.146°, 2.146°, 6.413°, 10.611°, and 14.697°, respectively.
In summary, we have developed a novel autostereoscopic display system that provides dynamic viewing zones with a reasonable display thickness to provide seamless stereoscopic views without abrupt change of 3D depth feeling at any eye position. This system can be applied to mobile devices, such as portable multimedia players, smartphones, and cellular phones. We have also extended it to multi-view 3D video services that provide different view images to match the viewer's position as he or she moves. These services include role-playing games, simulation games, and virtual reality.
Our proposed DLPB method, however, is limited by having only two beam-forming states. The more beam-forming states there are, the more viewing zones we can control, and the greater the degree of freedom the display provides. Currently, the need to track the position of the viewer means that our DLPB system operates for only a single viewer. In our future work, we plan to implement an ‘advanced dual layer parallax barrier’ method that uses a virtual PB layer to increase the number of beam-forming states without any structural changes to the current DLPB structure. We hope to extend it to a multi-viewer environment in the near future.
This research was supported by the Korea Communications Commission, under the ‘Development of Multi-view 3D Compatible UHDTV Broadcasting Technology’ support program supervised by the Korea Communications Agency.
Hyun Lee, Gi-Mun Um
Broadcasting System Research Department Electronics and Telecommunications Research Institute (ETRI)
Daejeon, South Korea
Hyun Lee was a researcher at Korea Telecom International from 1996 to 1999, when he joined the Broadcasting System Research Department at ETRI.
Gi-Mun Um has a PhD in electronic engineering and is a Principal Researcher at ETRI, which he joined in 2000. He worked at the Communications Research Center (Canada) as a visiting researcher from 2001 to 2002.
Electrical Engineering Department Korea Advanced Institute of Science and Technology (KAIST)
Daejeon, South Korea
Changick Kim was a senior member of the technical staff at Epson Research and Development Inc., Palo Alto, CA from 2000 to 2005. From 2005 to 2009, he was an associate professor in the School of Engineering at the Information and Communications University (Korea). He has been an associate professor at KAIST since March 2009.
1. S. Yi, H. Chae, S. Lee, Moving parallax barrier design for eye-tracking auto-stereoscopic displays, Proc. 3DTV Conf.: the True Vision—Capture, Transm. Display 3D Video, pp. 165-168, 2008. doi:10.1109/3DTV.2008.4547834
2. T. Peterka, R. L. Kooima, D. J. Sandin, A. Johnson, J. Leigh, T. A. DeFanti, Advances in the dynallax solid-state dynamic parallax barrier auto-stereoscopic visualization display system, IEEE Trans. Visualiz. Comput. Graph. 14, no. 3, pp. 487-499, 2008. doi:10.1109/TVCG.2007.70627