时间：７月１５日（周五）上午１０：００

地点：海韵行政楼Ｃ５０５报告厅。

报告内容: In the first part of the talk, I will describe my work on applying image segmentation for rigid shape representation/matching and its applications in object detection and tracking. This work is motivated by the observation that the boundaries of image segmentation capture distinctive global shape structures of images. As any single segmentation computed for an image can be unreliable, to address this problem, we developed a closed-form approximation to compute the average shape similarity between any two images by considering all possible segmentations for them. In the second part of my talk, I will describe a recent label fusion technique for multi-atlas based segmentation. In this work, we develop a new label fusion technique which results in more powerful label fusion algorithms. Our experiments on segmenting the hippocampus in magnetic resonance images (MRI) show that our method yields significant improvement over previous label fusion techniques.

个人简历: Hongzhi Wang received a PhD degree in computer science from Stevens Institute of Technology in 2008. His PhD work focused on perceptual organization and its usage in shape matching for object recognition. Since 2009, he has been a postdoctoral researcher at the Penn Image Computing and Science Lab, University of Pennsylvania. His current research interests are in computer vision, medical image analysis, perceptual organization, and machine learning.