Illumination invariance at a single pixel

This paper addresses the question of what can be said about the colors in images that is independent of illumination. We make two main assumptions: Firstly, the illumination can be characterized as Planckian (a realistic assumption for most real scenes). Secondly, the camera behaves as if it were equipped with narrow band sensors (true for a large number of cameras). An alternative set of assumptions (with broad-band sensors) leading to the same invariant expression is also presented to show the robustness of the invariant.The resulting physics-based method results in a transformation of the original color image to a grey-scale one which does not vary with illumination. Experiments demonstrate that the distribution of grey-scale invariants in an image is a reliable cue for illumination independent object recognition.