Self-similarity: a measure for benchmarking sewh-fragile image authentication techniques

This paper proposes the use of generalized finite automata (GFA) as a method for retrieving the image self-similarity to be used as a measure for benchmarking semi-fragile image authentication techniques. A comparison between PSNR as a conventional-method and the proposed self-similarity measure for several types of images was conducted. These comparisons show that the self-similarity measure has a better correlation with the image perceptual quality than the conventional PSNR. Furthermore, the self-similarity can be used to determine the maximum level of JPEG compression that should be used to benchmark a group of image authentication techniques