Illumination correction in digital images

Abstract

Illumination in an image often needs to be corrected or adjusted for several reasons. Recognition of images, such as those of faces, that are based on image data as patterns depends on having the lighting direction of the query image and that of the database images be aligned. In computational video, acquired images are often texture mapped to blend with computer generated polygonal surfaces. The shading of a polygonal surfaces changes as the light source direction relative to the composite object changes due, for instance, to the object's motion. We experiment with methods that can adjust the texture mapped image according to the illumination change. We test our concept on images of human faces