Color Image Compression Algorithm Using Self-Organizing Feature Map

A new color image compression algorithm using Kohonen\u27s self-organizing feature map is proposed. Our algorithm is an extension of color image compression algorithm proposed by Pei and Lo [IEEE Trans. Circuits Syst. Video Technol., 8: 191-205 (1998)]. N neurons are introduced in order to reduce a given full color image with 224 colors to an indexed color image with N colors. There are control parameters for the competitive learning among neurons in the self-organizing feature map algorithm. In our proposed algorithm, some of the control parameters, which are included in a neighboring function defined for neurons, are updated by taking relationship among neighboring neurons into account, though all control parameters are updated so as to decrease monotonically and exponentially with respect to each iteration step in Pei and Lo\u27s algorithm. The color palette obtained by the proposed algorithm is more robust as for control parameters than that by Pei and Lo\u27s algorithm