Face Recognition Under Expressions Occlusions and Pose Variations

Introduction of Project

We propose a completely unique geometric framework for analyzing 3D faces, with the precise goals of comparing, matching, and averaging their shapes. Here we represent facial surfaces by radial curves emanating from the nose tips and use elastic shape analysis of those curves to develop a Riemannian framework for analyzing shapes of full facial surfaces. This representation, along side the elastic Riemannian metric, seems natural for measuring facial deformations and is strong to challenges like large facial expressions (especially those with open mouths), large pose variations, missing parts, and partial occlusions thanks to glasses, hair, etc. This framework is shown to be promising from both – empirical and theoretical – perspectives. In terms of the empirical evaluation, our results match or improve the state-of-the-art methods on three prominent databases: FRGCv2, GavabDB, and Bosphorus, each posing a special sort of challenge. From a theoretical perspective, this framework allows for formal statistical inferences, like the estimation of missing facial parts using PCA on tangent spaces and computing average shapes.

Existing System

Existing face recognition techniques have demonstrated the potential of invariance to facial variations caused by illumination and have achieved high accuracy rates. to form the popularity process illumination invariant, phase congruency feature maps are used rather than intensity values because the input to the face recognition system. The feature selection process presented during this paper springs from the concept of modular spaces. Recognition techniques supported local regions have achieved high accuracy rates. Though the face images are affected thanks to variations like no uniform illumination, expressions and partial occlusions, facial variations are confined mostly to local regions. Modularizing the pictures would help to localize these variations, provided he modules created are sufficiently small. But during this process, great deal of dependencies among various neighboring pixels could be ignored. this will be countered by making the modules larger, but this is able to end in an improper localization of the facial variations.

PROPOSED SYSTEM

This paper proposes these problems within the framework of similarity matching. a completely unique perception inspired non-metric partial similarity measure is introduced, which is potentially useful in affect the concerned problems because it can help capturing the prominent partial similarities that are dominant in human perception. The effectiveness of the proposed method in handling large expressions, partial occlusions and other distortions is demonstrated on several well-known face databases.

The overall architecture of the proposed method is enable the capture of the partial similarity and therefore the integration of the spatial information, face images are partitioned into local facial regions (sub-blocks) initially . Then, all the sub-blocks are mapped into an SOM mathematical space to get a compact and robust representation, where the closest neighbor search is performed using the proposed partial distance measure, and therefore the training face image with the littlest partial distance to the probe face image is chosen to offer the ultimate identity.

Finally, the cropped face areas are processed by a histogram equalization algorithm to scale back the influence of possible illumination variations. The sizes of every cropped image within the AR. Then, the facial areas are cropped from the face image. Capture the relationships among quite two pixels; the info is projected into nonlinear higher dimensional spaces using the kernel method. this permits to capture the nonlinear relationships among the pixels within the modules. The experimentation procedure is conducted in two phases. within the first phase, the proposed feature selection strategy is tested on visual spectrum images obtained from AR database.