Title: Tree Filtering and Its Applications

Date & Time: Dec 27th (Fri) 10:00-11:30AM

Venue: 海韵教学楼208

Abstract: Edge-preserving filters like the bilateral filter have found widespread use in many computer vision and graphics tasks. Being non-linear, the brute force implementations of the bilateral filter are slow when the kernel is large.  This talk is focus on a new efficient edge-preserving filter named tree filter and its applications. The proposed tree filter explores a new paradigm for filtering an image: all pixels in the image are treated as nodes in a tree, and the filtered result of each pixel is determined by its position in the tree. Specifically, a minimum spanning tree (MST) is extracted from the original image with nodes being all the image pixels and edges being intensity/color differences between the nearest neighboring pixels. Given the MST, a standard efficient spatial smoothing filter like a recursive filter along its edges will be edge-aware as two dissimilar pixels will be automatically dragged away from each other in the MST. The effectiveness of the proposed tree filter has been experimentally verified for a variety of applications including stereo matching, image abstraction, edge extraction and texture-illumination decoupling.

Bio: Dr. Yang, Qingxiong is an assistant professor at City University of Hong Kong, China. He obtained his PhD degree in Electrical & Computer Engineering from University of Illinois at Urbana-Champaign (UIUC) in 2010 and BEng degree from University of Science & Technology of China (USTC) in 2004. Dr. Yang's research interests reside in computer vision. He won the best student paper award at MMSP 2010 and best demo at CVPR 2007. http://www.cs.cityu.edu.hk/~qiyang/