Neural Scene Segmentation by Oscillatory Correlation
Rept. of May 1996-Sep 1999
OHIO STATE UNIV COLUMBUS DEPT OF COMPUTER AND INFORMATION SCIENCE
Pagination or Media Count:
The segmentation of a visual scene into a set of coherent patterns objects is a fundamental aspect of perception, which underlies a variety of important tasks such as figureground segregation, and scene analysis. An innovative approach has been developed that uses neural oscillator networks to segment images. The framework, called oscillatory correlation, encodes the binding of pixels by phases of neural oscillators, resulting in a dynamical systems approach. The approach has been evaluated by computer simulations using both synthetic and real imagery, including intensity, range, motion, texture images. The results have been very competitive.