LDEye

One of my computer graphics courses at the Technical University Delft involved familiarizing myself with a 3D approximation/imposter technique called 'Layered-Depth Images' (or LDIs for short).

An LDI is basically a number of 2D images with additional depth information, all merged into one. The result is that each pixel is in fact a 'stack' of pixels (multiple 'layers of depth'). The inclusion of more depth information makes it possible to render LDIs at much closer ranges because they contain more so-called 'intra-object' occlusion information. The image is rendered by 'splatting' the pixel onto the screen.

Building the LDI structures can either be done by merging multiple ordinary depth images into a single LDI or by exposing a 3D computer model to a modified ray-tracer which detects multiple surface intersections per viewing-ray. The additional costs involved in rendering an LDI onto the screen remain low because the rendering process largely involves matrix computations which can be subdivided in such a way that successive layers of pixels can be positioned in the 3D world using only offsets from the base image plane.

For a more in-depth discussion of the techniques involved I would like to refer you to the paper that was written as part of the assignment. It suffices to say that LDIs can also be successfully used up to medium ranges while maintaining image quality at far cheaper rates than a normal polygon-based rendering would.