Colour constancy in natural images through colour naming and sensor sharpening

Abstract

Colour is derived from three physical properties: incident light, object reflectance and sensor sensitivities./nIncident light varies under natural conditions; hence, recovering scene illuminant is an important issue in com-/nputational colour. One way to deal with this problem under calibrated conditions is by following three steps, 1)/nbuilding a narrow-band sensor basis to accomplish the diagonal model, 2) building a feasible set of illuminants,/nand 3) defining criteria to select the best illuminant. In this work we focus on colour constancy for natural/nimages by introducing perceptual criteria in the first and third stages./nTo deal with the illuminant selection step, we hypothesize that basic colour categories can be used as anchor/ncategories to recover the best illuminant. These colour names are related to how the human visual system has/nevolved to encode relevant natural colour statistics. Therefore the recovered image provides the best represen-/ntation of the scene labelled with the basic colour terms. We demonstrate with several experiments how this/nselection criterion achieves current state-of-art results in computational colour constancy. In addition to this/nresult, we psychophysically prove that usual angular error used in colour constancy does not correlate with/nhuman preferences, and we propose a new perceptual colour constancy evaluation./nThe implementation of this selection criterion strongly relies on the use of a diagonal model for illuminant/nchange. Then, the second contribution focuses on building an appropriate narrow-band sensor basis to represent/nnatural images. We propose to use the spectral sharpening technique to compute a unique narrow-band basis/noptimized to represent a large set of natural reflectances under natural illuminants and given in the basis of hu-/nman cones. The proposed sensors allow predicting unique hues and the World colour Survey data independently/nof the illuminant by using a compact singularity function. Additionally, we studied different families of sharp/nsensors to minimize different perceptual measures. This study brought us to extend the spherical sampling/nprocedure from 3D to 6D./nSeveral research lines remain still open, such as, measuring the effects of using the computed sharp sen-/nsors on the category hypothesis; or inserting spatial contextual information to improve category hypothesis./nFinally,to explore how individual sensors can be adjusted to the colours in a scene.