The image-based representation of geometry is a well known concept in computer graphics. Due to z-buffering, the derivation of such representations using render-to-texture delivers only information of the closest fragments with respect to the virtual camera. Often, transparency-based visualization techniques, e.g., ghosted views, also require information of occluded fragments. These can be captured using multi-pass rendering techniques such as depthpeeling or stencil-routed A-buffers on a per-fragment basis. This video presents an additional rendering technique that enables the derivation of image-based representations on a per-object level within a single rendering pass. We use a dynamic 3D texture atlas that is parameterized on a per-frame basis. Prior to rasterization, the primitives are transformed to their respective position within the texture atlas, using vertex-displacement in screen space.
http://www.hpi.uni-potsdam.de/doellner/publications/year/2010/1048/TD10.html
Link to this comment:
All Comments (0)