Universal watermark estimation scheme based on error probabilities
Yoonki Choi, Kiyoharu Aizawa · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2001
In this paper, we propose a new watermark estimation technique based on error probability that gives a universal scheme for watermark estimation. This method treats the bits of extracted watermark as random variable and calculates the probability of certainty whether the image contains a certain watermark or not. Calculated probability value is a universal measure including the discrepancy of the number of bits of watermarks and shows how reliable the watermark is contained in an image. The results of experiments using still images and movies demonstrate that the probabilities based method is robust and efficient to estimate the extracted watermarks.