Abstract:
Digital watermarking of electronic data has become a popular research area over last
several years. The research work presented in this thesis blends a more practical
perspective of digital watermarking schemes against transformation/geometric attacks
with theoretical studies presented in past literature. Main objective of this dissertation
is to come up with an attack success/failure evaluation metric against chosen watermarking
schemes and with set of attacks. \Success" or \Failure" is defined as per given
criteria: If there is a legitimate owner for the image, success of image attack would be
determined if recovery of the watermark is denied given the fact that attacked image
lies within the perceptual range. Otherwise it would be depicted as a failure of image
attack.
In this thesis, limits that preserve perceptual quality of subjected images were analyzed
for different types of attacks. Both PSNR and Histogram Analysis provide a foundation
for meaningful measurement. We have shown that coupling such measurements
with a perceptual quality measurement provide meaningful values for effectiveness of a
watermarking scheme.
