Laird Robert Hocking

I'm fairly knowledgeable about image processing in particular and scientific computing in general. Right now I'm very interested in Chinese, but not yet very knowledgeable.

3D conversion refers to generating, given a single movie corresponding to one eye, a plausible movie for the other eye (forming a stereo 3D pair). A challenge arising in this is that areas behind occluding objects in original view may be visible in the new view. Therefore, something plausible has to be "painted in". The company I was working for had a very simple process for doing this, but it created some disturbing artefacts - for example, when inpainting a scene containing a picture frame in the background, the picture frame might suddenly bend sideways. To solve this, I first proved that their method was equivalent to a particular differential equation under a certain limit, enabling me to see what was going wrong, and then modify the algorithm so that the limiting differential equation did the right thing.