Image and Video Databases
Instructor: Vassilis Athitsos
The goal of this course is to study methods for comparing and identifying visual patterns. For example, these visual patterns can correspond to letters, hands, gestures, faces, or other types of objects or activities. The first topic we study is: what are meaningful ways to evaluate the similarity between visual patterns? How can we tell that a pattern A is more similar to a pattern B than to a pattern C? The course covers different similarity measures, including correlation, the Euclidean distance, the chamfer distance, the Hausdorff distance, dynamic time warping, the edit distance, and shape context matching. The second topic of this course is: how can we find efficiently the most similar match for a test pattern in a large database of images or video? The course covers different indexing methods, that can efficiently identify the best matches without having to exhaustively compare the test image or video to every database image or video.
Course web page:
Office: NH 309
Office hours: MW 3:00pm-4:30pm.