Feature difference analysis in video authentication system

Abstract

In most video authentication systems, the difference between features of the original and received videos is often used to decide the authenticity. In this paper, by employing mutual information to represent the similarity between the original and received video frames, we theoretically analyze the relationship between the feature difference and the video distortion. The relationship we derived is applied to estimate the maximum allowable feature difference in a video authentication system and to show how the feature difference varies with quantization step. Experimental results demonstrate that the derived relationship is reasonable and helpful for designing a robust video authentication system.