ILLUMINANT-TILT ESTIMATION FROM IMAGES OF ISOTROPIC TEXTURE

The authors present a new illuminant-tilt-angle estimator that works w ith isotropic textures. It has been developed from Kube and Pentland's frequency-domain model of images of three-dimensional texture, and is compared with Knill's spatial-domain estimator. The frequency and spa tial-domain theory behind each of the estimators is related via an alt ernative proof of the basic phenomena that both estimators exploit: th at is that the variance of the partial derivative of the image is at a maximum when the partial derivative is taken in the direction of the illuminant's tilt. Results obtained using both simulated and real text ures suggest that the frequency-domain estimator is more accurate.