A Progressive Non-discriminatory Intensity Equalization Algorithm for Face Analysis

Illumination plays a major role in the determination of image quality in an uncontrolled environment. Shadow casting, poor contrast and poor intensity are of particular interest in facial analysis. These problems are characterized by the nature of face image from uncontrolled environment. Several attempts have been made to enhance image quality; however, existing enhancement methods are limited in the specificity for facial expression analysis, thus resulting into non-uniform pixel brightness. This paper presents a novel method which is referred to as an intensity equalizer that handles objects of an image in singleton, pixelates the objects in this case a face image, employs the HSV color model for intensity value separation of every pixel, computes Gaussian probability density function and transforms every pixel with the best local minimum difference of mean intensity and variance of saturation across the object. Experimental result shows that the proposed model is invariant to shadow casting and gives moderate contrast on face image irrespective of input’s contrast class while preserving the global intensity information and enhancing local pixel information.