A Colour Quantization Method with Histogram Partition
by
Milen Todorow
South-West University, Blagoevgrad
Coauthors: Anton A. Penzov

Several colour quantization methods are analysed and described. Each algorithm has different computational complexity. The error from quantization increases with decreasing the computational price for execution of the algorithm. The effectiveness of each algorithm depends on the different color distribution in the color image for quantization.

A new approach for colour image quantization is proposed. The colour histogram of the image is generated and structured in a binary tree. Subsequently the colour histogram is partitioned into groups and for each group a representative colour is selected. Finally the original image is rewritten in accordance with the generated representative colours. By attempting to preserve the character of colour distribution in both original and rewritten image, the proposed method produces smaller error from quantization and runs faster than most colour quantization methods.

Date received: June 27, 2001