IRIS Research Staff

Venkat Rao Ayyagari
M.S. Student

(Click here to enlarge)

The non intrusive nature of facial data acquisition makes face recognition one of the most popular approaches for biometrics-based identity recognition. Mug shot matching, user verification and access control, and enhanced human computer interaction are all feasible if a robust face recognition system can be implemented.

We propose a technique that uses 3D geometric (point sets) face representations. The use of 3D point sets to represent human faces in lieu of 2D texture makes this method robust to changes in illumination and pose. The method first automatically registers facial point-sets through a criterion based on Gaussian force fields. The registration method defines a simple energy function, which is always differentiable and convex in a large neighborhood of the alignment parameters; allowing for the use of powerful standard optimization techniques. The new method overcomes the necessity of close initialization and converges in much less iterations as compared to the Iterative Closest Point algorithm (ICP). The use of an optimization method, the Fast Gauss Transform, allows a considerable reduction in the computational complexity of the registration algorithm. Recognition is then performed by using the robust similarity score generated by registering 3D point sets of faces.

The registration of range maps, and in general of 3D free from-surfaces, is an important task in many computer vision applications, including scene modeling and object recognition. For volumetric datasets several similarity measures were employed to align multi-modality datasets, which are mostly based on correlating intensity values. Currently, the focus of my work is to test the robustness of the algorithm developed for 3D registration based on Gaussian fields.

In this project, a survey on human face detection in color images and an implementation of automatic face detection based on skin color and face features are presented. The color information in YCbCr color space is combined with the multiple features such as eyes and mouth, to detect faces in arbitrary color images. This implementation consists of the following parts:  The first step in this program uses a skin model to determine the most likely skin regions in the image. A skin likelihood model is made based on the skin model of the different skin colors. Each likely skin region is then evaluated for different characteristics of a face. This program uses a template face to match to the segmented skin region, if the template face is close enough to matching the segmented region; the program determines that this region is a face. A rectangle is then placed around the face in the image. This process is repeated on each likely skin region.