Showing papers by "Charles R. Dyer published in 2008"

Image-Based Human Age Estimation by Manifold Learning and Locally Adjusted Robust Regression

Abstract: Estimating human age automatically via facial image analysis has lots of potential real-world applications, such as human computer interaction and multimedia communication. However, it is still a challenging problem for the existing computer vision systems to automatically and effectively estimate human ages. The aging process is determined by not only the person's gene, but also many external factors, such as health, living style, living location, and weather conditions. Males and females may also age differently. The current age estimation performance is still not good enough for practical use and more effort has to be put into this research direction. In this paper, we introduce the age manifold learning scheme for extracting face aging features and design a locally adjusted robust regressor for learning and prediction of human ages. The novel approach improves the age estimation accuracy significantly over all previous methods. The merit of the proposed approaches for image-based age estimation is shown by extensive experiments on a large internal age database and the public available FG-NET database.

Locally Adjusted Robust Regression for Human Age Estimation

Abstract: Automatic human age estimation has considerable potential applications in human computer interaction and multimedia communication. However, the age estimation problem is challenging. We design a locally adjusted robust regressor (LARR) for learning and prediction of human ages. The novel approach reduces the age estimation errors significantly over all previous methods. Experiments on two aging databases show the success of the proposed method for human aging estimation.

A Probabilistic Fusion Approach to human age prediction

Abstract: Human age prediction is useful for many applications. The age information could be used as a kind of semantic knowledge for multimedia content analysis and understanding. In this paper we propose a probabilistic fusion approach (PFA) that produces a high performance estimator for human age prediction. The PFA framework fuses a regressor and a classifier. We derive the predictor based on Bayespsila rule without the mutual independence assumption that is very common for traditional classifier combination methods. Using a sequential fusion strategy, the predictor reduces age estimation errors significantly. Experiments on the large UIUC-IFP-Y aging database and the FG-NET aging database show the merit of the proposed approach to human age prediction.

Head pose estimation: Classification or regression?

Abstract: Head pose estimation has many useful applications in practice. How to estimate the head pose automatically and robustly is still a challenging problem. In pose estimation, different pose angles can be used as regression values or viewed as different class labels. Thus a question is raised in our study: which is proper for pose estimation - classification or regression? We investigate representative classification and regression methods on the same problem to see any difference. A method that combines regression and classification approaches is also examined. Preliminary experiments show some interesting results which might prompt further exploration of related issues in pose estimation.

Easy as ABC? Facilitating Pictorial Communication via Semantically Enhanced Layout

Abstract: Pictorial communication systems convert natural language text into pictures to assist people with limited literacy. We define a novel and challenging problem: picture layout optimization. Given an input sentence, we seek the optimal way to lay out word icons such that the resulting picture best conveys the meaning of the input sentence. To this end, we propose a family of intuitive "ABC" layouts, which organize icons in three groups. We formalize layout optimization as a sequence labeling problem, employing conditional random fields as our machine learning method. Enabled by novel applications of semantic role labeling and syntactic parsing, our trained model makes layout predictions that agree well with human annotators. In addition, we conduct a user study to compare our ABC layout versus the standard linear layout. The study shows that our semantically enhanced layout is preferred by non-native speakers, suggesting it has the potential to be useful for people with other forms of limited literacy, too.

Face, Expression, and Iris Recognition Using Learning-based Approaches

Abstract: This thesis investigates the problem of facial image analysis. Human faces contain a lot of information that is useful for many applications. For instance, the face and iris are important biometric features for security applications. Facial activity analysis such as face expression recognition is helpful for perceptual user interfaces. Developing new methods to improve recognition performance is a major concern in this thesis. In approaching the recognition problem of facial image analysis, the key idea is to use learning-based methods whenever possible. For face recognition, we propose a face cyclograph representation to encode continuous views of faces, motivated by psychophysical studies on human object recognition. For face expression recognition, we apply a machine learning technique to solve the feature selection and classifier training problems simultaneously, even in the small sample case. Iris recognition has high recognition accuracy among biometric features, however, there are still some issues to address to make more practical use of the iris. One major problem is how to capture iris images automatically without user interaction, i.e., not asking users to adjust their eye positions. Towards this goal, a two-camera system consisting of a face camera and an iris camera is designed and implemented based on facial landmark detection. Another problem is iris localization. A new type of feature based on texture difference is incorporated into an objective function in addition to image gradient. By minimizing the objective function, the iris localization performance can be improved significantly. Finally, a method is proposed for iris encoding using a set of specially designed filters. These filters can take advantage of efficient integral image computation methods so that the filtering process is fast no matter how big the filters are.

Efficient Schemes for Monte Carlo Markov Chain Algorithms in Global Illumination

Abstract: Current MCMC algorithms are limited from achieving high rendering efficiency due to possibly high failure rates in caustics perturbations and stratified exploration of the image plane. In this paper we improve the MCMC approach significantly by introducing new lens perturbation and new path-generation methods. The new lens perturbation method simplifies the computation and control of caustics perturbation and can increase the perturbation success rate. The new path-generation methods aim to concentrate more computation on "high perceptual variance" regions and "hard-to-find-but-important" paths. We implement these schemes in the Population Monte Carlo Energy Redistribution framework to demonstrate the effectiveness of these improvements. In addition., we discuss how to add these new schemes into the Energy Redistribution Path Tracing and Metropolis Light Transport algorithms. Our results show that rendering efficiency is improved with these new schemes.