Partial Face Recognition Case Study

REVIEW ON ROBUST POINT SET MATCHING FOR PARTIAL FACE RECOGNITION
S.Varaprasad1, P.Jagadeesh2
Saveetha School of Engineering,
Saveetha University
Chennai

Abstract:
Over the past three decades, a number of face recognition methods have been proposed in computer vision, and most of them use holistic face images for person identification. In many real-world scenarios especially some unconstrained environments, human faces might be occluded by other objects, and it is difficult to obtain fully holistic face images for recognition. To address this, we propose a new partial face recognition approach to recognize persons of interest from their partial faces. Given a pair of gallery image and probe face patch, we first detect key points and extract

It do not require the presence of the eyes, face alignment or any other facial component within the image. In this case, we are not aware of a priori whether the input face is partial or holistic. Briefly, this approach is based on a Multi Key point Descriptor (MKD) representation which can be used for the probe image as well as the gallery dictionary. For each probe face, a Multitask sparse representation is to be learned, and thus for face detection, Sparse Representation-based Classification (SRC) approach is implemented. This method, we used to call is MKD-SRC. Shows the flowchart of the proposed method: The uniqueness of the proposed approach

State-of-the-art feature set matching techniques include Chui and Rangarajan presented feature set matching approach to align two point sets according to their geometry distribution by learning a non-affine transformation function embedded in a deterministic annealing process. Jian and Vemuri deployed Gaussian mixture models to represent the input point sets, and reformulated the problem of point set registration as the problem of aligning two Gaussian mixtures. Maier-Hein presented a convergent iterative closest point algorithm to accommodate anisotropic and inhomogeneous localization error. The above mentioned feature set matching approaches are not directly applicable for face recognition as most of them utilized only geometric or textural information of local features for matching. Differently, Li et al. [6] proposed a linear programming framework for feature matching, where both the geometric and textural features are employed. Their method is designed to match objects which are locally rigid and globally non-rigid, whereas human faces are globally rigid and locally non-rigid. The idea of feature set matching has also been exploited in face recognition. Used graph matching for face recognition. They first manually labeled face landmarks and then computed the face similarity based on local features around landmarks. In contrast our approach is a fully automatic approach, without