Histogram Matching by Moment Normalization

Abstract

A moment-based method for histogram matching is proposed in this paper to estimate the quasi-linear illumination change between two images which are composed of affine-transformed objects. For simplicity, the illumination change is modeled as a linear model with two unknown parameters to be estimated. Since histogram is invariant to object’s translation, rotation, and scaling simultaneously, it can be utilized to estimate the illumination relation between two images. B means of moment normalization, the moments of the two histograms can be normalized, by which the desired parameters can be estimated via the procedure of normalization. It has been shown that the proposed method can yield satisfactory estimations for an adequate range of image illumination change.