Video processing using additional information

Mahesh Ramachandran

University of Maryland

[Home Page]

Notice:  Hosted by Aravind Sundaresan. Note the non-usual day (Friday) and time (11am).

Date:  Friday September 04, 2009 at 11:00

Location:  EJ228 (SRI E building)  (Directions)

Present day computer vision problems involve processing and deriving information from long video sequences. Because of the tough challenges in vision, an important research focus has been to develop algorithms for video processing using additional measurements available along with image datasets. Video processing using additional information has gained interest because of the widespread availability of sensors like inertial measurement units, rangefinders etc. along with video cameras in several domains. In this talk, we primarily focus on an algorithm for fast structure from motion from image sequences using additional measurements about the vertical direction (gravity) and height of the camera. Using the assumed additional information, we derive a fast, scalable and parallelizable algorithm for structure from motion. We argue that this form of additional information is available in several different problem scenarios. We illustrate competitive results on several challenging datasets including the VIVID aerial dataset and the Google StreetView Research dataset. We will also briefly describe work on video stabilization, mosaicing and moving object detection in aerial sequences with and without additional information.

Mahesh Ramachandran is a Ph.D. candidate in the Dept of Electrical Engineering working with Prof. Rama Chellappa. He previously obtained his Masters degree from the same department, and his Bachelors degree (Electrical Engineering) from the Indian Institute of Technology Madras. In the summer of 2008, he worked in the StreetView group in Google Inc, Mountain View, CA. In the summer of 2007, he worked in the Visualization and Computer Vision Laboratory in GE Research, Schenectady, NY.

Please arrive at least 10 minutes early in order to sign in and be escorted to the conference room. SRI is located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in the visitors lot in front of Building E, and should follow the instructions by the lobby phone to be escorted to the meeting room. Detailed directions to SRI, as well as maps, are available from the Visiting AIC web page.