The holy grail of 3D is a display that lets you see 3D effects without glasses, filters or other viewing devices. These autostereoscopic displays—sometimes called glassless displays—are likely to catch on first with single-viewer handheld devices and computer monitors where it’s easy for you to adjust the viewing distance and angle so you are in the best place to see the 3D effects—called the sweet spot. Things get more complicated and more expensive when designing a screen for a group of people, all viewing it from different angles. At the moment autostereoscopic displays are used as monitors on twin-lens stereo cameras and camcorders so you can compose and review your images in 3D. They are also found in larger digital picture frames used to display 3D images and movies. It won’t be long before they become more widely used on other devices including computer monitors and TV screens. Although they are already available on such devices, their high cost limits them to commercial and scientific applications.

To understand how these displays work, it helps to begin by understanding that they can have hundreds of thousands or even millions of pixels. Each can be set to any one of millions of colors by mixing various brightnesses of just three colored subpixels—red, green and blue—usually arranged on the screen in repetitive vertical stripes.

By placing a thin optical layer (also called a lens plate or optical filter) over the screen, stripes of subpixels from different images can be "cast", or projected, towards each eye. There are two common types of optical layers—parallax barriers that block the light in certain directions and lenticular sheets that refract the light.

For images to be displayed in 3D on these displays, each stereo pair must first be interleaved (also called interzigging). In this process each image is digitally sliced vertically into strips one pixel or one subpixel wide. The strips from both images are then interleaved like a perfectly shuffled deck of cards so they alternate across the display. The odd numbered strips are all from the left photo in a stereo pair and cast to the left eye. Even numbered stripes are from the right photo and cast to the right eye. Since only half of the stripes from each image can be used if the original image width is to be retained, the other half are discarded so horizontal resolution drops by half, but vertical resolution is unchanged.

Parallax Barrier Displays

The optical layer on a parallax barrier display has parallel slits, typically 50 to 100 per inch, aligned with the underlying LCD screen in such a way that you see only one set of stripes with each eye. The left eye sees odd numbered stripes from the left images and the right eye sees even numbered stripes from the right image. The barrier with its slits is created with liquid crystals that can be switched back and forth between being opaque or transparent with a small change in voltage. By changing the voltage the screen can also be instantly switched between 2D and 3D modes. Unfortunately the parallax barrier blocks some of the light from the underlying screen so the screen needs an extra bright backlight and that requires more power, a drawback on battery-powered portable devices.

Lenticular Displays

On lenticular screens, the optical layer is covered by long cylindrical lenses—called lenticles—that look something like transparent corduroy. The focal point of the lenses fall on the underlying LCD panel and the lenses refract the screen’s light so you see the left and right stripes of pixels or subpixels with different eyes. These displays can be switched between 2D and 3D using software or hardware to cast the same image to both eyes or to change the optical characteristics of the lenticular lenses. Sharp and others have used a variant of this technology. An array of prisms on the screen’s surface, instead of lenses, directs the light from odd and even pixels to the left and right eye respectively. There is more on how this technology is also used to make 3D prints in Chapter 2.

Multiview Displays

A 3D image seen on an autostereoscopic display from a particular angle, is said to form a view. Depending on the device there can be as few as 2 views or as many as 9 with perhaps more to come by the time you read this. Autoseteroscopic displays for handheld devices and computer monitors are usually viewed by one person from a fixed position. This means the display needs only one view for both eyes for good 3D effects. As this technology is adapted to TV sets that we gather around, the number of views must be increased so viewers can see the 3D effects from different positions. Each view has an optimal observation angle from which you see the images at their maximum brightness, called the optimal observation spot. If you move your head left or right, you continue seeing the same view as long as you stay within a fan-shaped visibility zone of a view although the farther you move from the optimal viewing spot, the dimmer the image is. If you move far enough you enter the area covered by the next view because the central set of visibility zones, which is used by one viewer, is mirrored by additional views on either side for other viewers. As you walk past a multiview display, you pass through each of the views it projects and may be able to detect when you pass from one view to the next.

More views are theoretically better because more people in more positions can see the 3D effects. Unfortunately, as the number of views increases, the brightness and resolution of each view decreases although brightness can be improved by increasing the display’s backlighting and the resolution can be improved by increasing the resolution of the screen.

Autostereoscopic displays have issues that affect their quality:

• You may see the brightness changing and flipping as you move horizontally between viewing zones.
• A picket fence effect is a moire-like pattern caused by the gaps between sub-pixels being magnified by the lenticular layer.
• These displays can be switched between 2D and 3D and in 3D their horizontal resolution is only half of what it is in 2D. This is because the two images in a stereo pair are interleaved into the horizontal space usually occupied by one image.

Since head placement is so important when viewing autostereoscopic displays, eye tracking systems are being developed. They automatically adjust the displayed images so they follow your eyes as you move your head, eliminating the necessity for precise head-positioning.