An Interactive Color Image Quantization Method with Flexible and Subjective Color Selection Capability

Abstract

We propose a new method of color image quantization using flexible color space clustering to allow user to dynamically control the color selection. The system first uses spheres with radius D, which can be changed by users through a conversation with the system, to divide the L⁺u⁺v⁺ color space and allocate the same color to the pixels whose colors belong to the same sphere. The system then allows users to specify several areas to be emphasized and allocate more colors to those areas by using smaller spheres to divide the color space of those areas. Finally, to specify the colors to be emphasized accurately, the system provides four 'templates' to the users.
Several experimental results in which the images of 2²⁴ colors are quantized to 16-color images show the effectiveness of the proposed method.