The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical Engineering University of Southern California

Technical Report USC-SIPI-115

“The SCOOP Pyramid: An Object-Oriented Prototype of a Pyramid Architecture for Computer Vision”

by Herbert S. Barad

December 1987

This dissertation describes a working software prototype of a pyramid architecture, known as the SCOOP pyramid, to investigate its use and effectiveness in computer vision. The pyramid architecture is shown by simulation to be an effective architecture for a wide range of computer vision tasks from low level pixel-oriented operations to segmentation to high level symbolic operations. Results also show that processing overhead for a task can take more time than the task itself. This processing overhead includes the loading of convolution kernels and morphological look-up tables. An object-oriented methodology for modeling the individual processors and ports is used. The method for modeling and constructing the prototype is efficient and flexible. This encourages the fine-tuning of the architecture design. The prototype is used as a testbed for simulations of computer vision tasks and the results of these simulations are presented. The simulations include isolated tasks: convolution, edge detection, and segmentation. Also, a complete scenario to find bridges in LANDSAT aerial data is studied. This scenario is controlled by a simple knowledge base. The SCOOP architecture provides an environment to model architecture of arbitrary topology, complexity, and composition.

To download the report in PDF format click here: USC-SIPI-115.pdf (41.8Mb)