Simplified Mesh Generation from Renders

Scientific visualization presents data or real-world objects in 3D to visually verify and observe results. It has been used in everything from car manufacture to seismic analysis. The 3D models used frequently have far too much detail to be rendered at acceptable frame-rates on current computers, and while the speed of graphics hardware continually increases, so does the resolution of stored data. Quite often, the models displayed include very high resolution submodels that were originally intended to be displayed on their own, like including a fully detailed pressure gauge with all mechanical parts on a large engine. While many methods exist to reduce the complexity of the regular meshes such objects are built from, they all require detailed knowledge of mesh data. The only way to get this data is to integrate extended data export into the tool used to produce the mesh, a task which is difficult unless the software is self-written or comes with full source code. Presented here is a new method to generate a well-formed approximate mesh with color and normal information, based on just the ability to draw the object on screen. The resulting mesh contains no hidden or intersecting surfaces and hence is perfectly suited for simplification.

[1]  Peter Lindstrom,et al.  Fast and memory efficient polygonal simplification , 1998, Proceedings Visualization '98 (Cat. No.98CB36276).

[2]  Hugues Hoppe,et al.  Efficient implementation of progressive meshes , 1998, Comput. Graph..

[3]  Hugues Hoppe,et al.  Progressive meshes , 1996, SIGGRAPH.

[4]  Michael Garland,et al.  Surface simplification using quadric error metrics , 1997, SIGGRAPH.

[5]  Mongi A. Abidi,et al.  Mesh-based integration of range and color images , 2000, SPIE Defense + Commercial Sensing.