next up previous contents index
Next: 6.6 Character Recognition by Up: A Hierarchical Scheme Previous: 6.5.6 Results for Depth

6.5.7 Conclusions

The parallel simulation environment was written by Roberto Battiti [Battiti:90a]. Geoffrey Fox, Christof Koch, and Wojtek Furmanski contributed with many ideas and suggestions [Furmanski:88c].

A JPL group [Synnott:90a] also used the Mark III hypercube to find three-dimensional properties of planetary objects from the two-dimensional images returned from NASA's planetary missions, and from the Hubble Space Telescope. The hypercube was used in a simple parallel mode with each node assigned calculations for a subset of the image pixels, with no internodal communication required. The estimation uses an iterative linear least-squares approach where the data are the pixel brightness values in the images; and partials of theoretical models of these brightness values are computed for use in a square root formulation of the normal equations. The underlying three-dimensional model of the object consists of a triaxial ellipsoid overlaid with a spherical harmonic expansion to describe low- to mid-spatial frequency topographic or surface composition variations. The initial results were not followed through into production use for JPL missions, but this is likely to become an important application of parallel computing to image processing  from planetary missions.

Guy Robinson
Wed Mar 1 10:19:35 EST 1995