C. Cruz-Ramos, R. Reyes-Reyes, J. Mendoza-Noriega, M. Nakano-Miyatake, H. P?rez-Meana
In this paper, a verification criterion to reduce false positive and false negative error probability for image content authentication systems is proposed. Many image content authentication algorithms detect content altered region of images, but generally, both false positive and false negative error are considerably high and it makes impractical to use of these algorithms. The proposed verification criterion tries to eliminate blocks that are detected erroneously by authentication algorithm. The criterion uses neighboring concept based on 8-conectivity and it is evaluated using three image authentication algorithms based on semi-fragile watermarking technique. The false positive and false negative error probabilities for these three algorithms with the proposed criterion were calculated, when watermarked images were received content altered attack, such as photomontage, and content preserved attacks, such as JPEG compression and noise contamination. The experimental results show favorable performance of the proposed criterion, and the comparison results of three authentication algorithms suggest advantage and disadvantage of these three algorithms.