Inter-Camera Model Image Source Identification with Conditional Probability Features
Awa Wahab, Ats Ho, Shuangshuang Li · Surrey Research Insight Open Access (The University of Surrey) · 2012
In this paper, we propose a camera identification algorithm based on the conditional probability features (called CP features in this paper). Specifically, we report its performance for identification of image sources. Using four cameras of different models, we demonstrate that the CP features allow us to correctly identify the sources of 400 test images with an average accuracy of 99.50%. Additionally, the CP features based camera identification algorithm is also robust to cropping and compression. When the 400 images are cropped and JPEG compressed with QF=80 the average identification accuracy only slightly drops to 97.75%. These experimental results provide a good indication that CP features are promising new features for image forensics purposes.