Statistical Modeling of Visual Patterns

• Activity Number: 439
• Type: Other
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: ASA
• Author(s): Ying Nian Wu*+
• Companies: University of California, Los Angeles

Abstract

There are bewildering varieties of visual patterns in the natural scenes, including object patterns such as human figures, horses, birds, and texture patterns such as foliage, grass, river, as well as the articulate and complex motions of these patterns. How does biological visual system represent, learn, and recognize these patterns from image data collected by the retina? This problem appears to be effortlessly easy for biological vision, but it proves to be exceedingly difficult for computer vision. In this lecture, I shall review existing literature on statistical modeling of visual patterns. The lecture starts from the pattern theory pioneered and advocated by Grenandar and Mumford. It also includes our own past work on the models for object patterns and texture patterns.