Multispectral image compression for color reproduction; weighted KLT and adaptive quantization based on visual color perception

An inter-band decorrelation and quantization method of multispectral data are proposed, which are suitable for lossy compression of multispectral images used for high fidelity color reproduction. The proposed inter-band decorrelation method is a modified version of Karhunen-Loeve transform (KLT), called weighted KLT (WKLT), which is designed to minimize the color difference between the color images reconstructed from original and restored multispectral images. For quantization of WKLT coefficients, adaptive quantization (AQ) is introduced in order to equalize the partiality of the perceived error which is caused by the visual nonlinearity between the luminance and the perceived lightness of the color. Through the experiments using 16-band multispectral image of an oil painting, it is confirmed that WKLT followed by AQ reduces the average and the maximum color differences in L*a*b* color space in comparison with the conventional methods composed of KLT and linear quantization.