Alleviate the Effect of Illumination Variation in ICA Subspace

Abstract

Illumination variation brings in significant difficulties to pattern recognition and consequent robot operations. In this paper an illumination factor for real-world images taken under a wide variety of lighting conditions in ICA subspace was extracted. After normalizing with the combination coefficient of the illumination factors, all the images that originated from the same object but with different illumination intensities were easily classified.