Graphic Arts in Finland 35(2006)2 Color Naming and Computational Prediction from Natural Images

The article at hand is based on the master's thesis by one of the authors entitled "Color Naming and Computational Prediction from Natural Color Images " /21/. The work was conducted in the Laboratory of Media Technology at Helsinki University of Technology under the supervision of Professor Pirkko Oittinen and the instruction of Jyri Kivinen M.Sc.. The objective of the work was to develop a format for describing color information of digital images based on quantitative measures. The purpose of the proposed format was to serve the needs of the image production chain concerning various processes and related communication. The emphasis was on image retrieval and image selection. The principal aim was to support Finnish production teams. A method was developed for describing the dominant colors of natural images using linguistic expression. The method was based on the dominant color descriptor of the MPEG-7standard and it utilized ISCC-NBS color vocabulary for the linguistic conversion of the color values. Three experiments were conducted in order to gain information for constructing a modified version of the ISCC-NBS vocabulary and to test the method for naming dominant colors from images. The results proved that the dominant colors produced by the color naming method correspond well with human responses within limits of similarity among color names, even though some shortcomings were registered. The developed color vocabulary was, however, estimated to be too detailed for its purpose