The Imalab method for vision systems

Abstract.

We propose a method of constructing computer vision systems using a workbench based on a rich extensible toolbox and a general-purpose kernel. The toolbox provides access to an open set of libraries; the kernel provides incremental dynamic system construction and interactivity. This method makes it possible to quickly develop and test new algorithms, simplifies the use and reuse of existing program libraries, and allows to construct a variety of systems to meet particular requirements. Major strong points of our approach are: (1) Imalab is a single environment for different types of users who share the same basic code with different interfaces and tools. (2) New library modules are added quickly and easily, including libraries for scientific domains other than vision (e.g., robotics, Bayesian reasoning, automatic learning). (3) Different programming languages - C/C + + and several symbolic languages (Lisp, Prolog, Clips) - are tied together in a single system. We consider this an important advantage for the implementation of cognitive vision functionalities. (4) Automatic program generation simplifies the integration of libraries and makes the multilanguage feature work smoothly. (5) Efficiency: library code runs without overhead. The Imalab system has been in use for several years now, and we have started to distribute it.