3-D Faces Recognition With Missing Parts: A Review
Abstract— This paper presents a survey of face recognition which is a very important task of identifying human faces. Face
representations based on 3-D data are expected to be much more complicated to pose changes and illumination variations
than 2-D images, thus allowing accurate face recognition also in real-world applications with unconstrained acquisition.
FACE recognition using 3-D scans of the face has been recently developed as an alternative or complementary solution to
conventional 2-D face recognition approaches working on still images. A possible way to solve locally the problem of
missing data in 3-D face acquisition is to detect the absence of regions of the face and use the existing data to reconstruct the
missing parts. Traditional 3-D face modelling and recognition methods constrain human faces with either the Lambertian
assumption or the same assumption, resulting in suboptimal shape and texture models. This strategy is SIFT-based and
performs in a hybrid way that combines local analysis and holistic analysis, associating associating the keypoints between
two facial representations of the same subject. Global 3-D face representations for partial face matching have been given in a
limited number of works. A canonical representation of the face is that which exploits the isometry invariance of the face
surface to manage missing data obtained by randomly removing areas from frontal face scans. 3-D surface approximation
which considers spatial variability of specular and use reactions in face modelling and recognition testing. Instead of
evolving each test case individually, evolve all the test cases in a test suite at the same time. At the end, the best resulting test
suite is minimized. Personalized 3-D face models are estimated from a small number of face images under different lighting
conditions with a fixed viewpoint .Both shape and surface reactance properties are calculated locally through minimization
of the image differences between the original images and their estimations. Stochastic computational methods and
integrability enforcement are employed to handle the nonlinearity and inconsistency issues in the shape and reactance
parameter optimization processes to achieve valid results. Recently, pose variations in face recognition captured growing
interest from researchers in the fields of computer vision and pattern recognition. An effective strategy to handle pose
variations is to use the assistance of 3-D face models, because human heads are nonplanar 3-D objects so that viewpoint
changes take place in the 3-D space. Three dimensional model-based face recognition using personalized 3-D models
estimated from 2-D images has shown its promise. Such face recognition algorithms perform on a recognition by-synthesis
mechanism in the 2-D image spaces or in dimension reduction over shape and surface reactive parameters in the 3-D space.
Consequently, the performances of these techniques are greatly dependent on the accuracy of the estimated 3-D models.
Image-based face modelling considers facial textures (i.e., pixel intensities) as critical clues and estimates 3-D face shapes
and surface reactance properties by reversing image formation processes.