A novel approach for registration of 3D face images

This paper outlines a Nose tip localization technique in case of 2.5D as well as 3D meshes. No normalization process is applied and the process correctly localizes nose tip across any pose (including rotation in any direction in 3D space namely about x-axis, y-axis and z-axis).The present technique works by taking a facial image as input, and after which a thresholding process is applied to remove irrevelant details and finally the nose tip is detected using a maximum intensity technique as illustrated in Section III.c.1. Three dimensional face registration is a critical step in 3Dface recognition. In fact 3D faces still require being pose normalized and correctly registered for further face analysis.The 3D data may have different translation, rotation or scaling due to the controlled environment parameters such as the acquisition setup, device properties or due to uncontrolled conditions parameters such as the pose variations of the acquired subjects. In either case, the 3D shapes need to be aligned to each other and should be brought into a common coordinate frame before a comparison can be made. Registration is the alignment procedure of two similar shapes.Normally there is an importance of locating facial features e.g. lips, nose-tip which is required for face registration depending upon which alignment and consecutively registration has to be performed. The task of face registration is an issue due to the inherent elasticity present in human skin and the range of motion available to the human jaw. The aim of this paper is locating the facial points i.e. the nose tip. The present technique works by taking a facial image as input and after which a thresholding process is applied to remove irrelevant details and finally the nose tip is detected using a maximum intensity technique as is described in Section III.c.1 below. Experimented on nearly about 472 faces consisting of different poses (including rotation about x-axis, y-axis and z-axis) selected from the FRAV- D (Face Recognition and Artificial Vision Database), the present technique recognizes nose tip in 468 of the cases thus displaying a 99.15% of good nose tip localization.