A color classification algorithm

摘要

The color classification process requires that a color image bepartitioned into a set of uniform colored regions. A method of automaticclassification of a color picture, based on the advantages of ahigh-speed binary tree splitting algorithm and completed with a stoppingcriterion derived from color cluster detection methods, is presented.The histogram of distance of the current color point to the center ofthe investigated cluster is analyzed to stop the binary splittingalgorithm. The color space was divided and labeled accordingly to theclosest color cluster or region corresponding to a color of the colorset. A dynamic dividing method of the color space was used. Apost-processing algorithm makes it possible to filter out the unwantedtransition colors by absorption and erosion