An algorithm is proposed for 3D face recognition in the presence of varied facial expressions. It is based on combining the match scores from matching multiple overlapping regions around the nose. Experimental results are presented using the largest database employed to date in 3D face recognition studies, over 4,000 scans of 449 subjects. Results show substantial improvement over matching the shape of a single larger frontal face region. This is the first approach to use multiple overlapping regions around the nose to handle the problem of expression variation.

This work examines face recognition using normal intensity images, infrared images, three-dimensional shape, and combinations of these. We compare the performance improvement obtained by combining three-dimensional or infrared with normal intensity images (a "multimodal" approach) to the performance improvement obtained by using multiple intensity images (a "multisample" approach). Combining results from different types of imagery gives significantly higher recognition rates than are obtained by using a single intensity image. However, significantly higher recognition rates are also obtained by combining results from multiple intensity images. Overall, initial results indicate that, using an "eigen-face" recognition algorithm and weighted score fusion, multisample techniques can result in a performance increase comparable to that of multimodal techniques

We report on the largest experimental study to date in multimodal 2D+3D face recognition, involving 198 persons in the gallery and either 198 or 670 time-lapse probe images. PCA-based methods are used separately for each modality and match scores in the separate face spaces are combined for multimodal recognition. Major conclusions are: 1) 2D and 3D have similar recognition performance when considered individually, 2) combining 2D and 3D results using a simple weighting scheme outperforms either 2D or 3D alone, 3) combining results from two or more 2D images using a similar weighting scheme also outperforms a single 2D image, and 4) combined 2D+3D outperforms the multiimage 2D result. This is the first (so far, only) work to present such an experimental control to substantiate multimodal performance improvement.

Researchers have suggested that the ear may have advantages over the face for biometric recognition. Our previous experiments with ear and face recognition, using the standard principal component analysis approach, showed lower recognition performance using ear images. We report results of similar experiments on larger data sets that are more rigorously controlled for relative quality of face and ear images. We find that recognition performance is not significantly different between the face and the ear, for example, 70.5 percent versus 71.6 percent, respectively, in one experiment. We also find that multimodal recognition using both the ear and face results in statistically significant improvement over either individual biometric, for example, 90.9 percent in the analogous experiment.

Skin detection is an important process in many of computer vision algorithms. It usually is a process that starts at a pixel-level, and that involves a pre-process of colorspace transformation followed by a classification process. A colorspace transformation is assumed to increase separability between skin and non-skin classes, to increase similarity among different skin tones, and to bring a robust performance under varying illumination conditions, without any sound reasonings. In this work, we examine if the colorspace transformation does bring those benefits by measuring four separability measurements on a large dataset of 805 images with different skin tones and illumination. Surprising results indicate that most of the colorspace transformations do not bring the benefits which have been assumed.

This paper presents a method of evaluating unsupervised texture segmentation algorithms. The control scheme of texture segmentation has been conceptualized as two modular processes: (1) feature computation and (2) segmentation of homogeneous regions based on the feature values. Three feature extraction methods are considered: gray level co-occurrence matrix, Laws' texture energy and Gabor multi-channel filtering. Three segmentation algorithms are considered: fuzzy c-means clustering, square-error clustering and split-and-merge. A set of 35 real scene images with manually-specified ground truth was compiled. Performance is measured against ground truth on real images using region-based and pixel-based performance metrics.