Abstract
Primate visual systems support an elaborate specialization for processing color information. Concentrating on the hue component, we observe that, contrary to Mondrian-like assumptions, hue varies in a smooth manner for ecologically important natural imagery. To represent these smooth variations, and to support those information processing tasks that utilize hue, a piecewise smooth hue field is postulated. The geometry of hue-patch interactions is developed analogously to orientation-patch interactions in texture. The result is a model for long-range (horizontal) interactions in the color domain, the power of which is demonstrated on a number of examples. Implications for computer image processing, computer vision, visual neurophysiology and psychophysics are discussed.
Original language | English |
---|---|
Pages (from-to) | 753-771 |
Number of pages | 19 |
Journal | Neural Networks |
Volume | 17 |
Issue number | 5-6 |
DOIs | |
State | Published - 1 Jun 2004 |
Externally published | Yes |
Keywords
- Cytochrome oxidase blobs
- Horizontal connection
- Hue curvature
ASJC Scopus subject areas
- Cognitive Neuroscience
- Artificial Intelligence