Chromatic Statistical Landscape Features for Retrieval of Color Textured Images

This paper proposes a new texture description method named Chromatic Statistical Landscape Features(CSLF) for color images. The pro-hue pseudo intensity imageis extracted from a color image based on pro-hue pseudo function which measures the degree of a color containing the amount of the reference color component of hue. Besides prohue pseudo intensity image, the separate chromatic components saturation and intensity of a color image are also equivalent to two gray images. The graph of a gray image function is a rumpled surface appearing like a landscape, from which six novel texture feature curves are derived. In this way, eighteen feature curves can be obtained from the three gray images. The proposed Chromatic Statistical Landscape Features uses these eighteen feature curves to characterize the color texture and has been shown by systematic experiments to offer very high retrieval performance on the VisTex texture collection.