This week Sumod will give a very brief introduction to computer vision and talk about discrete optimization based approaches to solve some computer vision problems. Specifically, he will introduce the various areas in computer vision and its relations to various other fields. We will then learn how techniques like graph algorithms and dynamic programming help us in solving computer vision tasks such as image segmentation, object recognition, stereo and image restoration. We will understand how seemingly unrelated problems like shortest path between two points and the maximum amount of water that can flow in a network of pipes has structural similarities to some of these computer vision problems. Some of these discrete optimization techniques have proven to be remarkably good at various vision tasks. We will end with a few demos and compare the results of graph based algorithms versus other techniques.
Some knowledge of graph algorithms/optimization will be helpful. However, even if you have no background in graph algorithms but is interested to know more, please feel free to drop by.
1. If you need a refresher, please check out MIT OCW's (SMA5503) Introduction to Algorithms : Lectures 16, 17, 18.
2. Also helpful with a head start is the survey by Pedro Felzenszwalb and Ramin Zabih, "Dynamic Programming and Graph Algorithms in Computer Vision".