Rendering with coherent layers

For decades, animated cartoons and movie special effects have factored the rendering of a scene into layers that are updated independently and composed in the final display. We apply layer factorization to real-time computer graphics. The layers allow targeting of resources, whether the ink and paint artists of cartoons or the graphics pipeline as described here, to those parts of the scene that are most important. To take advantage of frame-to-frame coherence, we generalize layer factorization to apply to both dynamic geometric objects and terms of the shading model, introduce new ways to trade off fidelity for resource use in individual layers, and show how to compute warps that reuse renderings for multiple frames. We describe quantities, called fiducials, that measure the fidelity of approximations to the original image. Layer update rates, spatial resolution, and other quality parameters are determined by geometric, photometric, visibility, and sampling fiducials weighted by the ...

• Type: Conference
• Year: 1997
• Where: SIGGRAPH
• Authors: Jed Lengyel, John Snyder
• Keywords: Computer Graphics, Dynamic Geometric Objects, Layer Factorization, SIGGRAPH 1997