Has two polarising plates, as in a standard LCD display (link). Instead of a single layer of liquid crystal between the plates, there are several layers, each with its grid of electrodes. One of the layers, e.g. the one closest to the viewer, is chosen as the reference. For each pixel in this layer, ray-tracing software draws a path from the viewer's eye through each liquid crystal layer. The amount of rotation produced by each layer along this path is such that the total rotation is that needed to create the required brightness.
With n layers, there are n-1 degrees of freedom along the path from one eye. This provides redundancy such that the rotation can be correct for each eye. The mathematics is similar to that used in constructing CAT images. Because the brightness can be controlled independently for each eye, displaying 3D anaglyphs would be trivial.
Depending on the angle, between 1/2 and all of a pixel in the one layer will line up with a pixel in the next layer. This will produce a large decrease in maximum brightness, which could be partly offset by making the effective resolution less than the resolution of each layer.-- spidermother, Feb 22 2006 LCD display http://en.wikipedia.org/wiki/LCD_display [spidermother, Feb 22 2006] 3D LCD screen 3D_20LCD_20ScreenInspired by this. [spidermother, Feb 22 2006] Sliced_20cube_20dat...pression#1136339984 Although not at all related, it kind of reminded me of my (un-annotated <sob>) idea [xaviergisz, Feb 22 2006] //Is this just a more complicated way of doing Sharps parallax barrier ?// It sounds like it to me.-- zigness, Feb 22 2006 Not quite, although Sharp's parallax barrier is an excellent idea in itself. This started as an encrypted screen, once I worked out that the (linked) 3D LCD screen would have problems if you tried to view it from more than one direction at a time.
It was only halfway through writing that I realised you could exploit some of the redundancy to give different information to each eye.
[edit] //very similar idea// Actually, that sounds more like [espo]'s idea, in that it is a volumetric 3D display rather than a simulated point of view display.
But it differs from the parallax barrier in another way. Because of the ray-tracing from the eyes (obviously using well-baked tracking systems to know where the eyes are) a 3D image could be modified so that as you moved your head you would see it from different angles.-- spidermother, Feb 22 2006 [bigsleep] //there is no subtraction operation// Ah, but the arithmetic is modulo 180°. A 185° rotation is equivalent to a 5° rotation. And a 180° rotation is the same as no rotation at all. This is because (ignoring phase, which is unimportant here) polarised light is symmetrical about both the electric and magnetic field directions.
I was going to simplify the calculations by suggesting a black and white display, where each layer is capable of 0° or 90° of rotation. Thus an even number of 90° rotations is equivalent to no rotation (white), while an odd number is equivalent to 90° (black). This might be a good place to start.
Thanks for the CAT link. You made me laugh :-)
[edit] As I implied in the description, the physical pixels will rarely line up uniquely, one pixel per layer, but instead will often overlap. This will reduce control over darkness, as the pixel will not be evenly lit over its area, resulting in a rather yucky display. This is where use of smaller physical pixels than virtual pixels will help - the lack of control will only affect the physical pixels around the edge of a virtual pixel, and these could be masked out.
Also, the parallax interference effects of the masking (if present) around the pixels in each layer would be problematic; try looking at a scene through several closely spaced layers of flyscreen while moving your head around. Bleah.
Further simplification, such that the display only works for a very narrow range of viewing angles, would make these issues easier so solve. But that would kind of defeat the purpose...-- spidermother, Feb 22 2006 random, halfbakery