Geometry Compression

Michael Frank Deering: Geometry Compression


Company: Sun Microsystems, Inc.


The concept of  Geometry Compression (and Compressed Geometry) is the one sub-area of computer graphics that I am generally credited as the sole inventor of. My 1995 SIGGRAPH paper was the first of hundreds of papers that have since been published about the topic and related areas. The basic idea is that 3D geometry in computer graphics (vertices, normals, colors, texture coordinates, and connectivity information) can be subject to both lossless and lossy compression, just like many other representations of the real world (compression for images, compressions for audio). My goal wasn’t to produce the informational theoretical maximum compression (though I correctly predicted that many others would pursue this), but to produce a production compression format that could be implemented in both software and real-time hardware.

Because I had one Ph.D. student and two Masters students working on Geometry Compression tools at the time, I intentionally left out some of the more advanced software support I had already written and had running, in order to leave them room for their thesis (only one completed). For years I would wince every time a new paper would be published that basically only contained what I had already done (but not published); nowadays I believe that not anywhere as many papers would have been published in the field if I had included too much stuff in the first one. So if one counts the total number of papers published “as a good thing” this is OK then.

While most of the papers in the field were software techniques, my original concept worked both in hardware and software. The Elite3D included hardware support for decompressing geometry in real-time, but because of several hardware implementation glitches, the feature was not supported in the first two years of production.

A second expanded (but downward compatible) version of my hardware compressed geometry format was initially designed into both the Sun XVR-1000 and XVR-4000, but was not completed.


The main publications are the SIGGRAPH papers and course notes on geometry compression, plus the appendix in the Java 3D book.

Michael F. Deering. Geometry Compression. In Proceedings of ACM SIGGRAPH 1995.

pdf (94 KB) (presently minus all the images that show the results!)

Below are the notes  from my portion of the Geometry Compression courses taught in 1999 and 2000. The 2000 document is interesting, because it contains a public disclosure in compressed geometry level II (which was never productized, though most of the hardware design had been done).

Michael F. Deering. Hardware for Geometry Compression. Lecture in 3D Geometry Compression course, course organizer Gabriel Taubin, Proceedings of ACM SIGGRAPH 2000.

pdf (60 KB)

Michael F. Deering. Geometry Compression. Lecture in 3D Geometry Compression course, course organizer Jarek Rossignac, Proceedings of ACM SIGGRAPH 1999.

pdf (1 MB)

Michael F. Deering. Lecture in 3D Geometry Compression course, course organizer Gabriel Taubin, Proceedings of ACM SIGGRAPH 1998.