An alternative approach to image labeling using neural networks
T.A. Jamison, Robert J. Schalkoff · 2003
A solution of the image labeling problem using the emerging computational paradigm of neural nets is shown. An overview of neural network technology is provided and the labeling problem is formulated as a problem in symbolic constraint satisfaction. A neural network structure which embodies the labeling constraints is developed in detail. Experimental results and network convergence properties are analysed.>