Abstract: In this talk we will revisit some classical techniques from image processing and explore what is involved in translating them to the context of surfaces. We will show that by leveraging existing methodology from discrete differential geometry, it is often easy to extend the image-based techniques so that they can be used to edit signals on surfaces, including the surface geometry itself. In particular, we will focus on techniques in gradient-domain image-processing, shock filtering, and optical flow.

Speaker Bio: Kazhdan is considered a leader in image and geometry processing, especially surface reconstruction and shape analysis. He earned his bachelor’s (’97) from Harvard, and his master’s (’01) and Ph.D (’04) in computer science from Princeton University.

Host: Keenan Crane

For Appointments: Chris Downey (cdowney@andrew.cmu.edu)