Histogram Generation from the HSV Color Space Using Saturation Projection

Abstract

In this chapter, we make an in-depth analysis of the visual properties of the HSV (Hue, Saturation, Intensity Value) color space and describe a novel histogram generation technique for color feature extraction. In this new approach, we extract a pixel feature by choosing relative weights of hue and intensity based on the saturation value of the pixel. The histogram retains a perceptually smooth color transition between its adjacent components that enables us to do a window-based smoothing of feature vectors for the purpose of effective retrieval of similar images from very large databases. The results have been compared with a standard histogram generated from the RGB color space and also with a histogram similar to that used in the QBIC system (Niblack et al., 1993).