Improvements for multispectral image compression for color reproducibility with preservation to spectral accuracy

Abstract

This paper proposes a method for multispectral image compression that makes multispectral image applicable for both color reproduction and spectral image analysis. In the methods previously proposed for colorimetric error reduction, the weighting matrices were incorporated to Karhunen-Loeve transform (KLT) as the spectral transform for compression, which account for the color matching functions of human observer and the viewing illuminant. However, the improvement for color reproducibility is on the cost of spectral accuracy degradation. In this paper, we show the reduction of colorimetric error and the preservation of spectral accuracy is a tradeoff that can be controlled by adding a diagonal matrix to the weighting matrix of KLT. Experimental results using the proposed method integrated with JPEG2000 for a 16-band multispectral image show that spectral accuracy is improved without substantial colorimetric loss by the addition of diagonal matrix if the magnitude of the diagonal matrix is chosen appropriately.