3-D face recognition

[摘要] ENGLISH ABSTRACT:In recent years face recognition has been a focus of intensive research but has still notachieved its full potential, mainly due to the limited abilities of existing systems to copewith varying pose and illumination. The most popular techniques to overcome this problemare the use of 3-D models or stereo information as this provides a system with the necessaryinformation about the human face to ensure good recognition performance on faces withlargely varying poses.In this thesis we present a novel approach to view-invariant face recognition that utilizesstereo information extracted from calibrated stereo image pairs. The method is invariantof scaling, rotation and variations in illumination. For each of the training image pairs anumber of facial feature points are located in both images using Gabor wavelets. Fromthis, along with the camera calibration information, a sparse 3-D mesh of the face can beconstructed. This mesh is then stored along with the Gabor wavelet coefficients at eachfeature point, resulting in a model that contains both the geometric information of theface as well as its texture, described by the wavelet coefficients. The recognition is thenconducted by filtering the test image pair with a Gabor filter bank, projecting the storedmodels feature points onto the image pairs and comparing the Gabor coefficients from thefiltered image pairs with the ones stored in the model. The fit is optimised by rotating andtranslating the 3-D mesh. With this method reliable recognition results were obtained ona database with large variations in pose and illumination.

[发布机构] Stellenbosch University