Plastic Surgery Face Recognition Using Photometric Based Illumination Normalization Techniques and Gabor Faces.

Samar, Ahmed Shaaban Madian; Ibrahim, Rehab M.

doi:10.21608/bfemu.2020.106745

Plastic Surgery Face Recognition Using Photometric Based Illumination Normalization Techniques and Gabor Faces.

Document Type : Research Studies

Authors

¹ Head of Electronics and Communications Engineering Department., Faculty of Engineering., Mansoura University., Mansoura., Egypt.

² Electronics and Communications Engineering Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt

10.21608/bfemu.2020.106745

Abstract

Arising apparent great challenge to face recognition which is not much considered before is plastic surgery. The availability of advanced technological, at ever decreasing coasts, makes facial plastic surgery increasingly affordable and thus widespread. Previous research and results on plastic surgery were unable to provide acceptable levels of identification performances using current state of art face recognition techniques. This paper proposes an effective combination scheme that is able to improve plastic surgery verification/ identification rates. As the appearance of a face image is a severely affected by illumination conditions that will hinder the automatic face recognition process. A study of the effect of different photometric illumination techniques is carried on plastic surgery images in this paper. The effect of illumination is effectively reduced, and the contrast is enhanced by histogram equalization. The resulted face image is not only reduced illumination effect but also preserved edges and details that will facilitate further face recognition task. By combining photometric illumination techniques with Gabor representation of the images which have proven themselves to be a powerful tool for facial feature extraction and robust face recognition [1] and a group of holistic face recognition techniques gives us the advantages of no need of any prior information of light sources or 3D shape, nor many training samples needed thus can be directly applied to single training image per person condition. Experimental results carried on plastic surgery database, an ad-hoc database reported by Singh et al [2] shows a significant improvement in recognition rates than that reported before at different cases of plastic surgery.

Keywords

Main Subjects