Computer Science Colloquia
Wednesday, March 20, 2013
Connelly Barnes, Postdoctoral Research Scientist, Adobe
Host: Jason Lawrence
3:30 PM, Rice Hall, Room 130 (auditorium), followed by a reception in Rice Hall Fourth Floor Atrium (west end)
Mid-level Image Editing with PatchMatch
My talk presents my work about making it easier for people to create digital content. To achieve this goal, my strategy is to exploit statistics of the natural world to make more intuitive and higher level editing tools. In this talk, I will focus in particular on creative image editing and describe the PatchMatch algorithm and its extensions as well as the applications that it enables. This algorithm addresses a core problem of finding dense correspondences that is of longstanding interest in the vision and graphics communities. It uses stochastic search to accelerate nearest neighbor matching in images by one to two orders of magnitude due to underlying properties of the statistics of natural images. Applications include image editing, video tapestries, image analysis, and image stitching and morphing. I will conclude with a discussion of the impact on the research community and future goals I have.
*Mr. Barnes is a faculty candidate for the Department of Computer Science.
Bio: Connelly Barnes is a postdoctoral research scientist at Adobe. He received his Ph.D. from Princeton University in 2011. He works in computer graphics, and focuses on creation and manipulation of image and video. His technology has been incorporated into Adobe Photoshop headline features Content-Aware Fill and Move. He enjoys the outdoors, traveling, and snow skiing.