Genetic optimisation of illumination compensation methods in cascade for face recognition

Abstract

Face detection and recognition depend strongly on illumination conditions. In this reported work, genetic algorithms are used to optimise parameters of the modified local normalisation and self-quotient image methods in cascade for illumination compensation to improve face recognition. The main novelty of the proposed method is that it applies to non-homogeneous as well as homogeneous illumination conditions. The results are compared to those of the best illumination compensation methods published in the literature, obtaining 100% recognition on faces with non-homogeneous illumination and significantly better results than other methods with homogeneous illumination.

Illumination normalisation method using Kolmogorov‐Nagumo‐based statistics for face recognition

“…This is especially important in most real applications where illumination cannot be controlled. Face recognition is important in many applications such as security systems, access control, human–machine interfaces, gesture‐based computer interfaces, automatic driver monitoring, biometric identification and video search [1]. Face images are significantly changed by lighting conditions, which may cause performance degradation in both face detection and face recognition [2].…”

“…airports, banks, malls, train stations, public buildings. Therefore, it is highly desirable to develop new methods for face recognition that can compensate the illumination changes [1].…”

Gender Classification From Face Images Using Mutual Information and Feature Fusion

“…The high variability in the appearance of the face directly affects their detection and classification. Automatic classification of gender from face images has a wide range of possible applications, ranging from human-computer interaction to applications in real-time electronic marketing in retail stores (Shan 2012;Bekios-Calfa et al 2011;Chu et al 2010;Perez et al 2010a).…”

Illumination compensation method using edge‐weakening guided image filter for face recognition

“…Furthermore, these methods require either knowledge about the light source or large amounts of training data, which are not readily available in real-world situations. Among the existing illumination compensation methods, the self-quotient image (SQI) [2] has been demonstrated to be quite efficient and effective [3].…”