| 概要(Abstract)： |
Most of the pictures we see nowadays (press, fashion, politics, social networks) usually come from digital cameras. Those images can easily be created but also altered and manipulated with more and more sophisticated image editing softwares. Digital image forensics is an emerging research field that attempt to detect these image falsifications and restore trust towards digital imagery, with obvious applications in the press, justice and police investigations.
During this talk, I summarize some current image editing techniques available in most good image editing softwares. I also introduce the main forgery procedures using these tools, such as copy-past from an image to another, copy-move, but also color and geometric editing as well as some supervised tools (liquefy, inpainting, seam carving, ...). Then, I give an overview of a large set of digital image forensics methods to detect image forgeries. These techniques can take advantage of the image compression format. Some others focus on the image consistency on noise, color and geometrical clues. Finally, I present some counter forensics pretending to make undetectable falsifications.
|