Multilinear image analysis for facial recognition

Abstract

Natural images are rhe composite conseqirence of mulriple facrors related to scene structure, illumination, and imaging. For facial images, the factors include different facial geometries, expressions, heud poses, and lighting conditi[~ns. We upply multilinear algebra, rhe algebra of higherorder tensors, to obrain a parsimonious representation of facial image ensembles which separates these facfors. Our represenratioti, called TensorFaces, yields impmved facial recognirion rues relative to standard eigenfaces.

Fibre Centred Tensor Faces

“…Multilinear analysis was introduced to the computer vision community by Vasilescou and Terzopoulos [1][2][3][4]. Multilinear data represent the natural extension from scalars (0-D tensors), through vectors (1D tensor) and matrices (2D tensors) to general ndimensional data.…”

“…In their original tensorface work [1] Vasilescou and Terzopoulos proposed finding the optimal set of identity vectors for each set of non-identity parameters (expression, lighting, viewpoint). The optimal set of these weights that best matches the target are considered the best match.…”

Personalized Modeling of Facial Action Unit Intensity

“…In other words, we seek to find two low-dimensional spaces: the identity space and the AU-intensity-specific space, which, together, generate the observed facial features. Various approaches that address this task have been proposed [19,4,20,21]. These are based on the tensor representation of different factors (i.e., identity, pose, illumination, and expression), decoupling of which is attained by means of multilinear generalizations of Singular Value Decomposition (SVD).…”

Automatic Frequency Band Selection for Illumination Robust Face Recognition

“…The second approach tries to remove the illumination variation either by an image transformation or by synthesizing a new image [3], [4], [5], [6]. Finally in the third approach, illumination variation is learned and modeled in a suitable subspace [7], [8], [9], [10], [11]. Besides these solutions, in [12] near-infrared lighting is proposed to have illumination invariant capture conditions.…”

“…Face recognition under varying lighting has attracted significant attention and there have been many solutions proposed for this problem to provide illumination robust face recognition [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12]. These solutions can be classified as: invariant features, canonical forms, and variation modeling [3].…”