Visual location of license plates by vector quantization

The paper describes a methodology based on Vector Quantization (VQ) to support visual vehicle identification: license plate location is the specific task involved by VQ-based image coding. Using VQ yields superior picture compression for archival purposes and supports effective location at the same time. VQ encoding can give some hints about the contents of image regions; such information is exploited to enhance location performance. Training the VQ system by examples gives the advantage of adaptive on-field tuning. The approach has been tested in a real industrial application and included satisfactorily in an ATS for vehicle identification.