Face to Iris Area Ratio as a feature for children detection in digital forensics applications

Abstract

This paper presents a new geometrical feature based on iris geometry for automatic children detection in digital images. The proposed feature is based on simple geometrical measures of face elements and especially iris. Iris outer boundary is approximated as a circle and the proposed feature is defined using the iris area and the distance of iris from mouth line. This approach is compared with two similar geometrical features. Experimental results show the superiority of the proposed feature over the existing ones. The proposed feature is more robust compared to the existing features that use geometrical information from human faces and gives better classification results, useful to digital forensics applications.