New approaches for colour histogram adaptation in face tracking tasks

This paper proposes two new schemes to adapt histogram-based description of colour that aim to cope with colour change due to varying illumination conditions. Colour histograms have been extensively used because of their worth in easily describing object colour but change in colour quickly degrades their performance. Changes in nature of light may cause this problem which becomes extremely critical in tasks involving relative motion between objects and the visual system, as can be the case of face tracking based on skin colour. We extend the use of colour histograms in these situations by their adaptation. Results encourage the use of these new schemes and the description of colour using non-parametric models.