LivingKnowledge goal is to bring a new quality into search and knowledge management technology for more concise, complete and contextualised search results.
This paper, co-written by F. Uccheddu, A. De Rosa, A. Piva and M. Barni, proposes a proper experimental methodology to make a deeper-than-usual analysis of the performance of resampling detectors. The suggested framework has been applied to two of the most popular resampling detectors proposed so far. It has been presented at 18th European Signal Processing Conference on August 23-27 2010 in Aalborg, Denmark.
The assessment of the practical performance of forensic methods for the detection of resampling operations on digital images is a difficult task requiring a careful experimental analysis. Unfortunately, the experimental analysis reported in multimedia forensics papers is often statistically insufficient, or at least not usable to predict the performance of the proposed systems at scales needed for practical deployment. This paper attempts to move a first step to fill this gap in a twofold way. First of all a proper experimental framework is proposed to analyze the performance of resampling detectors. Then the suggested methodology is applied to evaluate the performance of two of the most significant resampling detectors proposed so far, providing a deeper-than-usual analysis of their behavior under different working conditions.