This paper describes a clusterization procedure based on a set of new transformations for HSL color space. The clusterization is applied for color layer extraction from the large blue print documents overdrawn with color pens and pencils. The process of color layer extraction requires a segmentation and a classification of segmented area to recompose each color layer. Overlapped colors are identified based on extraction of primary and secondary clusters. The new HSL transformations are based on the modification of lightness function in a standard HSL space. The modified HSL color space remains device dependent space, but the RGB regulate domain is not transformed into a HSL regulate domain. The clusterization procedure is based on the Mahalanobis distance in new HSL space. A 3D visualization procedure is used for illustration the efficiency of the clusterization process.
Gabriel Marcu, Satoshi Abe, "Color Clusterization using Modified HSL Space" in Proc. IS&T 4th Color and Imaging Conf., 1996, pp 151 - 155, https://doi.org/10.2352/CIC.1996.4.1.art00040