Application of the Karhunen-Loeve transform for natural color images analysis

Abstract

The current systems of image capture are based on red, green and blue (R,G,B) principles. But, this model of capturing color images is different from the human visual system. So, to obtain a representation which approaches the human system, one uses the intensity and chrominance space. As this representation is non-linear, it introduces color instability. Thus, to analyze natural color images, we use the Karhunen-Loeve space, which allows a large decorrelation of the color components, a high definition of colors and an increased compression ratio for very homogeneous images.