Part of Advances in Neural Information Processing Systems 5 (NIPS 1992)
Christof Koch, Binnal Mathur, Shih-Chii Liu, John Harris, Jin Luo, Massimo Sivilotti
We describe two successfully working, analog VLSI vision circuits that move beyond pixel-based early vision algorithms. One circuit, implementing the dynamic wires model, provides for dedicated lines of communication among groups of pixels that share a common property. The chip uses the dynamic wires model to compute the arclength of visual contours. Another circuit labels all points inside a given contour with one voltage and all other with another volt(cid:173) age. Its behavior is very robust, since small breaks in contours are automatically sealed, providing for Figure-Ground segregation in a noisy environment. Both chips are implemented using networks of resistors and switches and represent a step towards object level processing since a single voltage value encodes the property of an ensemble of pixels.
1 CONTOUR-LENGTH CHIP
Contour length computation is useful for further processing such as structural saliency (Shaashua and Ullman, 1988), which is thought to be an important stage before object recognition. This computation is impossible on an analog chip if we