ROTATION AND SCALE CHANGE INVARIANT POINT PATTERN RELAXATION MATCHINGBY THE HOPFIELD NEURAL-NETWORK

Relaxation matching is one of the most relevant methods for image matc hing. The original relaxation matching technique using point patterns is sensitive to rotations and scale changes. We improve the original p oint. pattern relaxation matching technique to be invariant to rotatio ns and scale changes. A method that makes the Hopfield neural network perform this matching process is discussed. An advantage of this is th at the relaxation matching process can be performed in real time with the neural network's massively parallel capability to process informat ion. Experimental results with large simulated images demonstrate the effectiveness and feasibility of the method to perform point pattern r elaxation matching invariant to rotations and scale changes and the me thod to perform this matching by the Hopfield neural network. In addit ion, we show that the method presented can be tolerant to small random error. (C) 1997 Society of Photo-Optical Instrumentation Engineers.