Title :

New Results in Image and Video Completion

Speaker :

Dr. Chiman Kwan

Chief Technology Officer

Signal Processing, Inc., Rockville, USA

Venue :

Room 215, William M. W. Mong Engineering Building, CUHK

Date :

Oct 17, 2011, Monday
2:30 PM - 3:30 PM

Abstract :

In this seminar, we present some interesting results in matrix completion of images and videos with missing data. First, since many conventional matrix completion approaches do not work well for actual images, which do not satisfy the low rank condition, we developed a new and high performance matrix completion algorithm for various types of real images, including LADAR and electro-optical (EO) images. Our algorithm can recover randomly missing pixels with different missing rates (50%, 75%, 90%, 95%, 99%, and 99.9%). Extensive simulations using 18 LADAR images and 23 EO images clearly demonstrated that our missing pixel recovery performance is excellent even in the extreme cases (99% and 99.9%). Second, we have developed a novel matrix completion algorithm for dealing with missing blocks in images. Many real images have been used to evaluate our algorithm. Our algorithm outperforms other methods and can also handle missing pixels covered by texts. Third, we developed novel matrix and tensor completion algorithms for video completion. Comparison with a well-known method called boundary matching algorithm (BMA) in the literature demonstrated that our algorithms indeed yielded better performance.

Biography :

Dr. Chiman Kwan received his B.S. degree in electronics from the Chinese University of Hong Kong in 1988 and M.S. and Ph.D. degrees in electrical engineering from the University of Texas at Arlington in 1989 and 1993, respectively.

From April 1991 to February 1994, he worked in the Beam Instrumentation Department of the SSC (Superconducting Super Collider Laboratory) in Dallas, Texas, where he was heavily involved in the modeling, simulation and design of modern digital controllers and signal processing algorithms for the beam control and synchronization system. He received an invention award for his work at SSC. Between March 1994 and June 1995, he joined the Automation and Robotics Research Institute in Fort Worth, where he applied intelligent control methods such as neural networks and fuzzy logic to the control of power systems, robots, and motors. Between July 1995 and March 2006, he was the Vice President of Intelligent Automation, Inc. in Rockville, Maryland. He served as Principal Investigator/Program Manager for more than sixty five different projects, with total funding exceeding 20 million dollars. Currently, he is the Chief Technology Officer of Signal Processing, Inc., leading research and development efforts in chemical agent detection, image/speech processing, and fault diagnostics and prognostics.

Dr. Kwan’s has published more than 50 papers in archival journals and has had 120 additional papers published in major conference proceedings.

    **************************************** ALL ARE WELCOME ****************************************

Enquiries: Ms. Winnie Wong or Prof. Wen J. Li, Department of Mechanical and Automation Engineering, CUHK at 2609 8337. *MAE Series (2011-12) is contained in the World-Wide Web home page at http://www3.mae.cuhk.edu.hk/maeseminars.php#mae.

Back