Implementation and Evaluation of the IUA C++ Class Library on the Connection Machine 1
|
Abstract:
It is always of interest how portable a language is, especially when it has been designed for a specific machine and application. It is also of interest how well languages specific to a machine can be used for porting another language onto this machine. In this project, the Image Understanding Architecture (IUA) C++ Class Library (ICCL) was ported to the Connection Machine (CM) with PARIS. Design tradeoffs are described. The portability of the ICCL, and the suitability of PARIS for this task are investigated. We explored performance bottlenecks in the transport of the ICCL on three different machines by running several algorithms, written in the ICCL, and comparing execution times. Also, a new vision algorithm was developed, which was optimized for speed in the ICCL. The suitability of the ICCL for this task and design criteria are described below.
Citations
| 87 | Extracting straight lines – Burns, Hansen, et al. - 1986 |
| 7 | The Effect of Spatial Discretization on the Magnitude and Direction Response of Simple Differential Edge Operators on a Step Edge – Kitchen, Malin - 1989 |
| 6 | The Class Library for the IUA: Tutorial – Burrill - 100 |
| 4 | Parallel Dense Depth from Motion on the Image Understanding – Weems - 1993 |
| 2 | Real Time Feature Extraction: A Fast Line Finder for Vision-Guided Robot Navigation – Kahn, Kitchen, et al. - 1990 |
| 1 | The Image Understanding Architechture – Weems, Levitan, et al. - 1989 |
| 1 | An Integrated Understanding Benchmark: Recognition of a 2 1/2 D 'Mobile'." CmpSci – Weems, Riseman, et al. - 1988 |
