Useful information plane on pattern classification

Abstract

This paper presents pattern classification by using the useful information plane of a gray level image through a pattern anlysis scheme. For gray level images, the original image is converted into two useful information planes, called differentiation and histogram planes. They are individually applied to the presented pattern analysis scheme, which possesses simultaneously invariant property under shift, rotation and scaling when there is one object in the input. A set of human face photographs is used for experiments. Results with the respective information plane are compared and show that the performance by the histogram plane is better than that by the others.