Representing Smooth Plane Curves for Recognition: Implications for Figure-Ground Reversal
Abstract
A rcprcscntntion of smooth pldnc curves for recognition is proposed. The basic rcprcscntation IS a linked list of four primitive shapes, called “codons”, which arc invariant under rot&ltions. translations and unifonn scaling. Psychophysical observations regarding the pcrccption of figure-ground reversals are prcscntcd to suggest that a similar representation could bc used by the human visual system. I lN’I’ROI~UCTlON A vision system somctimcs must compute its initial descriptions of a shape without bcnclit of context. Yet these descriptions should highly constrain the set of posslblc matches with memory if they arc to be useful for recognition. For cxaniplc. though one cannot rcasonnbly predict the contents of figure 1 prior to seeing it, the shapes arc readily rccogniLcd. This simple demonstration implies the cxistcncc of context-indcpcndcnt rules tllat provide shape descriptions which can bc used to initi‘itc the recognition process. Such rules for smooth plant curves are
Cite
Text
Hoffman and Richards. "Representing Smooth Plane Curves for Recognition: Implications for Figure-Ground Reversal." AAAI Conference on Artificial Intelligence, 1982.Markdown
[Hoffman and Richards. "Representing Smooth Plane Curves for Recognition: Implications for Figure-Ground Reversal." AAAI Conference on Artificial Intelligence, 1982.](https://mlanthology.org/aaai/1982/hoffman1982aaai-representing/)BibTeX
@inproceedings{hoffman1982aaai-representing,
title = {{Representing Smooth Plane Curves for Recognition: Implications for Figure-Ground Reversal}},
author = {Hoffman, D. D. and Richards, W. A.},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {1982},
pages = {5-8},
url = {https://mlanthology.org/aaai/1982/hoffman1982aaai-representing/}
}