M. Al-Athamneh¹, F. Kurugollu¹, D. Crookes¹ and M. Farid², ¹Queen’s University Belfast, United Kingdom and ²University of Derby, United Kingdom

This paper proposes a simple but yet an effective new method for the problem of digital video camera identification. It is known that after an exposure time of 0.15 seconds, the green channel is the noisiest of the three RGB colour channels [5]. Based on this observation, the digital camera pattern noise reference, which is extracted using only the green channel of the frames and is called Green-channel Photo Response Non-Uniformity (G-PRNU), is exploited as a fingerprint of the camera. The green channels are first resized to a standard frame size (512x512) using bilinear interpolation. Then the camera fingerprint is obtained by a wavelet based denoising filter described in [4] and averaged over the frames. 2-D correlation coefficient is used in the detection test. This method has been evaluated using 290 video sequences taken by four consumer digital video cameras and two mobile phones. The results show G-PRNU has potential to be a reliable technique in digital video camera identification, and gives better results than PRNU.

PRNU, G-PRNU, Video Forensics.