3D transformation matrix compression and decompression

Compressing and decompressing techniques for transformation matrices 3D computer graphics systems use to animate objects achieve high compression ratios by taking advantage of common characteristics of homogenous 3D transformation matrices. The techniques use a bitmap to encode information on locations of ones and zeros of the matrix--bypassing the penchant of compilers to represent such information as high-precision numbers. Since most video game processors and display hardware are constrained by their resolutions and since an original transformation matrix often stores data that is more accurate than necessary, the techniques convert some real numbers in the matrix (e.g., those within the range of -1 and 1) into integers by scaling them by a constant. The resulting compressed matrices occupy much less storage space than their non-compressed counterparts, and can be efficiently decompressed in real time for use in interactive real time 3D animations.

for bigger explanation and full example http://www.patentstorm.us/patents/6591019/description.html