Approximated Chi-square distance for histogram matching in facial image analysis: Face and expression recognition

Chi-square (χ
²
) distance is a useful metric for histogram matching in computer vision problems. However, it has more computational cost than Euclidean distance. In this paper, a new distance formulation is proposed to reduce the computational cost of χ
²
. In the proposed distance, the denominator of formula can be calculated in the feature extraction phase. Consequently, the computational cost of feature matching phase is considerably reduced. The proposed distance metric is evaluated using LBP, HOG, and POEM histogram features on different face datasets (including: CK+, JAFFE, and Yale) for face and facial expression recognition. The experimental results show that the proposed distance is 2.5 times faster than χ
²
with nearly the same accuracy.