Self-dependent 3D face rotational alignment using the nose region

Abstract

One of the challenging issues for 3D face recognition is face alignment. Many alignment algorithms are computationally expensive, making them unsuitable for real-time biometrics, or not robust enough to detect large variations in pose. In this work, a novel algorithm for 3D face rotational alignment is proposed, that uses the nose region. After preprocessing and nose region identification, alignment is performed by apply ing two energy functions to the nose footprint, identified as the largest filled region in the inverted depth map. These functions are minimised using Simulated Annealing and the Levenberg Marqurdt algorithm. The energy minimisation and segmenta tion procedures continue iteratively until a stopping criterion is met. The method has been applied to images from the Face Recognition Grand Challenge (FRGC) v2 dataset and the con sistency of its alignment has been verified using the iterative closest point (ICP) algorithm. As a self-dependent algorithm, it does not require a pre-aligned image as a reference and also has a high computational speed, approximately three times faster than the brute force ICP technique.

Using nasal curves matching for expression robust 3D nose recognition

“…To do this, the principal curvature and shape index (SI) are computed and the SI is scaled so that its maximum and minimum values are exactly +1 and -1, receptively. The face's convex regions are found by thresholding the SI to produce a binary image, using -1 < SI < − 5 8 [3,1,5,7]. The largest connected component is detected and its boundary is smoothed by dilating with a disk structuring element.…”

“…1(d). This approach to nose region cropping results in fewer redundant regions than the approach of [5] and is much faster than that of [4] which uses level set based contours.…”

“…The SI is again thresholded to extract the largest convex regions from the cropped nose region. Then, the nose region's depth map, Z n , is inverted and the largest connected region is located [5]. The resulting binary image is multiplied by the convex region to refine it and remove noisy regions.…”

“…It is also one of the parts of the face that is least prone to occlusions caused by hair and scarves [5]. Indeed, it is very difficult to deliberately occlude the nose region without attracting suspicion [7].…”

Expression robust 3D face recognition by matching multi-component local shape descriptors on the nasal and adjoining cheek regions

“…For example, the nasal region, which is more consistent over universal expressions and also invariant to the majority of occlusions caused by hair, hands and scarves [1][2][3][4][5], is widely used in many face recognition algorithms. Mian et al [3] found the forehead is also relatively invariant under basic expressions.…”