To appear in: Proceedings of the 4th International Conference on Artificial Intelligence: Methodology, Systems, Applications. Albena, Bulgaria, September 1990. J.J.Gibson argued that the study of 'ecological optics' is central to the understanding of vision. Gibson's own development of his idea was too qualitative to be useful in building machine vision systems; on the other hand, workers in computer vision have not yet accepted the implications of his view and developed fully explicit models of optic array structure. I suggest that two currently active developments, fractal modelling of images and simulated neural networks, might provide starting points for the quantitative ecological optics that is required for progress in computer vision.
This paper is not available online