Information theoretic analysis of the capacity of multi channel digital image watermarking

Abstract

Watermarking channel capacity is defined as the highest rate in bits per image of information that can be embedded with arbitrarily low probability of detection error under a given distortion between original and watermarked image. It is analogous to channel capacity defined in information theory. A mathematical method of decision-making called Game Theory, in which a competitive situation involving two parties is analyzed to determine the optimal solution for an interested party is very popular in information theory. Our approach for finding the amount of information that can be embedded in a digital image uses some principles of this theory. The two parties involved here are the information hider and the attacker. The information hider and attacker have a limit on the maximum amount of distortion each can introduce. The concept of game theory can be applied to watermarking as, information hider tries to embed as much information as he can constrained to maximum allowable distortion and the attacker tries to remove the embedded information by attacking the watermarked data constrained to maximum allowable distortion. Experiments are conducted by considering the above scenario. The results and conclusions are discussed in later sections of this thesis.