Performance impact of ordinal ranking on content fingerprinting

Content fingerprinting provides a compact representation of multimedia objects for copy identification. This paper analyzes the impact of the ordinal-ranking based feature encoding on the performance of content fingerprinting. Expressions are derived for the identification performance of a fingerprinting system with and without ordinal ranking. The analysis indicates that when the number of features is moderately large, ordinal ranking can improve the robustness of the fingerprinting system to large distortions of the features and significantly increase the probability of detection. These results enhance understandings of ordinal ranking and provide design guidelines for choosing different system parameters to achieve a desired identification accuracy.