Object Recognition by Contour Extraction

Joseph J. Lim : Computer Science/Applied Mathematics

Mentor: Professor Jitendra Malik, EECS

Object recognition is a major unsolved problem in Computer Vision. The main goal is for computers to detect and to recognize objects in the given images and videos. In this project, contours will be used as a new descriptor. "Contours" are defined as sets of segments that can provide more information than just a single segment or a random set of segments. In the past, not much work has been done using contours, because extracting contours from images was a difficult task. However, with a much improved contour extraction method developed in the Berkeley Vision Group, Joseph Lim and his mentor expect to contribute to a new method that will benefit the entire field of Computer Vision.