Institute of Information Science, Academia Sinica



Press Ctrl+P to print from browser


An Overview of Image Dehazing

  • LecturerMing-Sui (Amy) Lee (Department of Computer Science and Information Engineering)
    Host: Dr. Hong-Yuan Mark Liao
  • Time2010-02-26 (Fri.) 10:00 – 11:00
  • LocationAuditorium 106 at new IIS Building
Abstract Fog, haze and smoke that degrade the scene images are caused by the airlight which is the ambient light absorbed and scattered in the atmosphere before it reaches the camera. The further the object away from the camera, the fainter the object in the scene. This degradation causes images to lose contrast and color precision which is annoying for many applications in computer vision such as feature detection, object recognition, photometric analysis, surveillance system, etc. Therefore, image haze removal has attracted a growing interest recently. Since the haze effect highly depends on the distance between the object in the scene and the camera, how to estimate an accurate depth map is a challenging issue. Many different approaches were proposed including single image dehazing, multiple image dehazing, and some algorithms with additional information and user input. With the developments of dehazing techniques, both amateur and commercial photographers benefit a lot for various purposes. In this talk, recent researches related to image dehazing will be reviewed and some possible future directions will also be discussed. Short Biography Ming-Sui Lee received her B.S. degree in mathematical sciences from National Cheng-Chi University, Taiwan, in 1999. Then she received the M.S. degree in electrical engineering from the University of California, Los Angeles (UCLA) and the Ph.D. degree in electrical engineering from the University of Southern California (USC) in 2002 and 2006, respectively. Now she serves as an assistant professor at the Department of Computer Science and Information Engineering, National Taiwan University. Her research interests include image/video mosaicking, compressed-domain image/video processing, and super resolution techniques.