Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Guodong Guo


Family relationship analysis has many potential applications, ranging from homeland security through to image search and social activity analysis. In our work, we present five computational problems for family relationship analysis in face photos. Studying these challenging problems is important and useful for semantic image understanding and social context extraction. In our study, the familial traits are learned from pairs of salient local facial parts using discriminative approaches. It is motivated by human perception studies on kinship recognition and the existence of familial traits through genetic inheritance. Second, kinship verification is performed on a pair of faces by integrating the familial traits based on confidence measures. Then, the generation recognition and specific family relationship recognition are explored. Finally, the separation of family and non-family group photos is studied based on a decision that combines multiple pair-wise kinship detections. An image database consisting of both family and non-family group photos is collected, and labeled at different levels of details. Experiments are performed on the database for all five tasks, based on different representations of the facial parts. Preliminary results show that the proposed problems can be addressed with a reasonably good performance. Our encouraging results may inspire more effort from the computer vision and image processing research community.