Monday, February 22, 2010
Computer Studies Bldg. Room 209
Micah (Kimo) Johnson
Massachusetts Institute of Technolgy
Making and Breaking Fake Images
When making fake images, whether they are composites of photographs or purely computer-generated, it is difficult to achieve realism across various attributes. Color, noise and illumination are all important and adjusting these attributes independently requires expertise. In this talk, I will present my research on both sides of this problem, from detecting inconsistencies in digital forgeries to improving the appearance of seamless composites. By working on both problems, I have gained an appreciation for the factors that contribute to realism in imagery. These factors can be exploited for both new forensic tools and computer graphics techniques.
Bio: Kimo Johnson is a joint postdoc at MIT and Adobe Systems. His primary research interests are related to creating and detecting fake images. In this work, he draws inspiration from both computer vision and graphics, but he is also interested in general image and audio analysis, computational photography, art analysis and music theory. He holds undergraduate degrees in Mathematics and Music Theory from the University of New Hampshire and an A.M. in Electro-Acoustic Music from Dartmouth College. He received his Ph.D. in Computer Science from Dartmouth College where he was advised by Hany Farid.
Refreshments will be provided at 10:45