Detecting corresponding segments across images using synchronizable pulse-coupled neural networks

Computational models of locally connected networks of synchronizable neural oscillators-notably pulse-coupled neural networks (PCNN) and locally excitatory globally inhibitory oscillator networks (LEGION)-have been applied to image segmentation. We report on research that explores a simple 2-layer PCNN-like network for determining corresponding segments in two images. If the two images are frames in a video sequence this can be used for motion detection and, thus, for motion-based segmentation. More generally, it can be used for finding specific objects in fresh views of a previously imaged scene. The proposed algorithm is called bidirectional gated block-matching