SPIKE_4096: A neural integrated circuit for image segmentation
An image segmentation algorithm, based on Pulse-Coupled Neural Networks, was implemented in silicon. We aimed at simplifying neuron hardware implementation while maintaining segmentation efficiency. Some algorithmic tricks have then been added, improving the results. The main components of the underlying neuron architecture are a single 8 bits register, a simple incrementer, and some glue logic. A prototype, using a data flow architecture, implementing a 64x64 neuron array, and based on a 0.2 Μm CMOS SOI technology, will be released in 1998. A 64x64 segmentation is expected in less than 50 Μs.
Unable to display preview. Download preview PDF.
- 1.Ambiehl, Y., Pulse-Coupled Neural Networks for image segmentation (in French). Technical Report, Sup’Aéro-ONERA/CERT, Sept. 1997.Google Scholar
- 3.Clastres, X., Freyss, L., Samuelides, M., Tarr, G. L., Dynamic segmentation of satellite images using Pulse-Coupled Neural Networks. ICANN’95, Paris, Oct. 1995.Google Scholar
- 4.Clastres, X., Samuelides, M., Pulse coupled neurons and image segmentation (in French). Technical Report ONERA nℴl/35566.000/DERI, Dec. 1996.Google Scholar
- 5.Eckhorn, R., Reitboeck, H. J., Arndt, M., Dicke, P., Feature linking via synchronization among distributed assemblies: simulation and results from cat visual cortex. Neural Computation, 2, 293–307, 1990.Google Scholar
- 7.Pelloie, J. L., Faynot, O., Raynaud, et al., A Scalable Technology for Three Successive Generations: 0.18, 0.13, and 0.1 Μm for Low-Voltage and Low-Power Applications. Proceedings 1996 International SOI Conference, p.118, Oct. 1996.Google Scholar