We consider pose recognition, localization and segmentation of the whole body as well as body parts in images. This research is a fundamental step toward video-based human motion analysis that have been intensively studied recently. Pose recognition, localization and segmentation in a still image or image sequences are challenging problems due to the variability of human body shapes and poses. Our goal is to develop a hybrid human representation (see the right figure) and the corresponding processing to assemble three tasks into one integrated framework, where spatial, shape, and temporal priors are involved and fused at both part and whole levels. The proposed research is deeply inspired and motivated by shape representation theories in cognitive psychology and recent biological vision as well as the recent advancements in the field computer vision.

Related Publications

• C. Chen and G. Fan, "Hybrid Pose Representation for Integrated Pose Recognition, Localization and Segmentation," in Proc. IEEE International Conference on Computer Vision and Pattern Recognition, Anchorage, Alaska, June 23-26, 2008.

• C. Chen and G. Fan, "What Can We Learn from Biological Vision Studies for Human Motion Segmentation", in Proc. International Symposium on Visual Computing, Lake Tahoe, NV, Nov. 11-13, 2006, also LNCS, Vol. 4292, Editors: G. Bebis, Springer, 2006.

• C. Chen and G. Fan, "Perception Principles Guided Video Segmentation", in Proc. IEEE International Workshop on Multimedia Signal Processing, Oct. 2008, Shanghai, China.