The architectural feasibility of a generic computer vision system composed of heterogeneous processors is examined. The study focusses on issues such as the nature of the heterogeneous processors functional capabilities, network realization tradeoffs, and common module interfaces. The generic computer vision system is designed to meet the demands imposed by future industrial vision applications, including gray scale processing, real-time operation and low cost. Our research takes advantage of recent advances in high speed image and signal processing architectures, and on-going research in processor networking strategies. The program objectives are 1) the identification of common processing modules for future computer vision system applications, and 2) the identification of feasible inter-module networking to permit computer vision system reconfiguration for diverse industrial applications. The research will have vast implications in the transfer of computer vision technology to a broad range of applications.