Subject : Robust digital image watermarking
 
 Students :     Vinayak Dangui vdangui@stanford.edu
 
                     Nicolas Schifano schifano@stanford.edu
 
 
 Project proposal :
 
                     Watermarking consists in the insertion of a key within a digital file (MP3, image...), and is used for copyright purposes.  This technique(also sometimes referred to as steganography), has been developing since the seventies, when it was investigated by IBM.  This project has for objectives to implement a scheme of robust and invisible watermarking, and study the effect of various attacks on its detection, to ascertain its robustness.
 
 
                     For the implementation of the watermark, we will use a spread spectrum technique.  The invisibility of the watermark will be ensured by taking into account the psychoperceptual effects(i.e implantation in the middle range frequencies, etc...).  The detection will be implemented with an objective of very high discrimination (verified by submitting a set of the same image watermarked with different keys, and observing the peak in response corresponding to the correct key).
 
 
                     Regarding the attacks, we intend to study the following ones : addition of an additional watermark, addition of gaussian noise, low pass filtering, non-linear filtering, JPEG compression, cropping + rotation, printing + scanning, and collusion (the most dangerous and likely attack possible).  The performance of the watermark against each of these attack will be decided by observing if the detector can still discriminate the correct key in a set of various keys applied to the same image, who underwent the same attack.  Some of those attacks might require to use the original image (not watermarked) for detection purposes.
  
                     If the time constraints allow it, we intend to include forward-error correction in the watermarking insertion and decoding systems.