You are here: FIDIS Interactive > PRNUCompare > 

FIDIS Interactive

PRNUCompare - Photo Response Non-Uniformity (PRNU) Compare

In a forensic context it is interesting to be able to identify the source camera, based on the videos or images it produces. In order to do so, certain unique characteristics are needed that distinguish one camera from the others. To identify the source camera, the Photo Response Non-Uniformity (PRNU) that acts as a device signature is used as the unique characteristic. This PRNU presents itself as a visually invisible pattern that is present in all images or videos a camera produces. However, the degree in which this pattern is available depends a.o. on the illumination; in well illuminated segments of an image the PRNU can be more reliably estimated than in segments with low intensities. With the use of filters, these characteristic patterns can be extracted from images, or from individual frames from videos. 

In practice, device identification works as follows. We will speak in general of images, as videos can be thought of a sequence of multiple images. 

  • The pattern is extracted from a (questioned) image (or multiple questioned images), and denoted as the ‘natural pattern’,
  • Reference patterns’ are generated by making flatfield images: images that do not contain scene content and have an (approximate) uniform illumination. Alternatively, normal images with scene content may also be used, but the former is recommended. The PRNU patterns are then extracted from each of these flatfield images, to generate a reference pattern for each camera.
  • If the natural pattern and one of the reference patterns have a high degree of similarity, it is an indication that the natural image originates from this camera. Of course, the natural pattern and the reference patterns need to be the same size.

These steps are very easy to accomplish in the FIDIS/NFI PRNUCompare program. If desired, the algorithm can be changed from the default Gaussian filter to a wavelet based extraction method (View -> Advanced settings). This works better, especially for videos, but is slower. Further help can be found in the Help file (Press ‘F1’ once the program is running).

Images

Extracted PRNU pattern (variation in greylevels in colour)
Extracted PRNU pattern (variation in greylevels in colour)
Real image taken with camera
Real image taken with camera
Extracted pattern from real image
Extracted pattern from real image
PRNU pattern extracted from test images
PRNU pattern extracted from test images

Links / Resources

  • The program is freely available from the website: https://sourceforge.net/projects/prnucompare/ .
  • Natural videos and Reference videos are available from http://www.youtube.com/user/gbus1982.The filenames indicate the origin of the video as <cameramodel>.<cameranumber> - codec settings.avi , e.g.: 12.6 - natural wmv9 q50 640x480 : camera model 12 (Logitech Communicate STX), cameranumber 6, natural video encoded with the wmv9 codec in resolution 640x480 with quality setting 50. These videos can be downloaded with keepvid.com .