Region duplication or region cloning is a very common practice of image tampering, where a continuous portion of pixels in an image are pasted to a different location to conceal undesirable objects or contents in the original image. In this work, we describe a new detection method based on matching image SIFT features. The robustness of the SIFT features with regards to local transforms renders this method able to detect general region duplications under geometric and illumination distortions with efficient computation.

• References:
  
  
  1. X. Pan and S. Lyu, “Region Duplication Detection Using Image Feature Matching”, IEEE Transactions on Information Forensics and Security (TIFS), 5(4):857-867, 2010. [Impact Factor: 2.338] [In top 10 accessed articles of IEEE TIFS March 2011] ( Featured in VerticalNews on Jan. 11, 2011. )
     [PDF | BibTex]
     
  
  
  2. X. Pan and S. Lyu, “Detecting Image Region Duplication using SIFT Features”, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Dallas, TX, March 14 – 19, 2010. (Oral Presentation)
     [PDF | BibTex]
     
  
  
  3. X. Pan and S. Lyu, “Exposing Doctored Images by Detecting Duplicated and Distorted Regions”, Technical Report SUNYACS-09-02, Department of Computer Science, SUNY Albany, November 2009.
     [PDF | BibTex]
     
  
  
  
• Experiments:

• 	Region Duplication under Geometric Transform using Photoshop We show seven representing cases of tampered images using region duplication, along with their detections using our method as filled regions. The five cases correspond to (1) simple copy-move, (2) scaling of the duplicated region with factor 1.2, (3) rotation of the duplicated region of 60°, (4) illumination adjustment, (5) a shear transform (combination of rotation and scaling) of the duplicated region, (6) reflection transform, and (7) perspective transform. We create these forgeries manually using Photoshop, and take particular care in selecting source regions and finessing the duplicated region to achieve convincing splicing effects. The original image is courtesy of Alin C Popescu in the Image Science Group led by Prof. Hany Farid at Dartmouth College. Experimental Results can be found here.
  	Region Duplication using State-of-the-art Computer Graphics Technology The forgeries in this experiment are created using a coordinate-based approach that optimally render the transition between duplicated region and the surroundings at the target location, leading to highly convincing tampering results. For these images, our method can recover fairly accurately the duplicated regions, and correctly accounts for the geometrical and illumination distortions The original and the tampered images are courtesy of the authors of "Coordinates for Instant Image Cloning" - SIGGRAPH 2009. Experimental Results can be found here.
  	Region Duplication using Commercial Software This forgery is created with the Smart Fill tool in the Image Doctor 2 software (Alien Skin Software). The unpublished algorithm used to create this forgery is more sophisticated: instead of using a continuous duplicated region of relatively large size, smaller regions containing mostly textures (sometimes the selected region has size of less than 20 pixels) are combined and arranged to cover larger region at the target location. This makes the visual detection of the duplicated regions significantly more difficult. It also poses as a challenge to our method, especially due to that the smaller identical regions provide less reliable keypoint matchings. However, the detection result can still provide considerable clue to drawn an inspector's attention for scrutiny. The original and the tampered images are courtesy of Alien Skin Software LLC. Experimental Results can be found here.
  	Real-life Forgery on Nationally Recognized Media As another set of test, we run our detection method on an image that has appeared on the front pages of several internationally important newspapers including The Los Angeles Times, The Financial Times, and The Chicago Tribune and several major news Web sites including The New York Times early this year. Shortly after this image was published, doubts have been raised that it has been digitally altered, a fact later confirmed by inspection of photography experts and the appearance of another photography that was taken under similar condition. We use our detection method to analyze this image, and are able to find two major region duplications in this image. Consistent with the expert analysis by visual inspection, our method provides an objective and quantitative evaluation. The original and the tampered images are courtesy of The New York Times. Experimental Results can be found here.
  	Video Tampering using Region Duplication The forgeries in this video are created using a coordinate-based approach that optimally render the transition between duplicated region and the surroundings at the target location, leading to highly convincing tampering results. The detected duplicated regions in the tampered video are highlighted. Our method can recover fairly accurately the duplicated regions, and correctly accounts for illumination and slight geometric adjustments. The original and the tampered videos are courtesy of the authors of "Coordinates for Instant Image Cloning" - SIGGRAPH 2009. Experimental Results can be found here.

Back to Research Homepage