Loading...

Quantization-unaware double JPEG compression detection

Taimori, A ; Sharif University of Technology

1054 Viewed
  1. Type of Document: Article
  2. DOI: 10.1007/s10851-015-0602-z
  3. Publisher: Springer New York LLC
  4. Abstract:
  5. The current double JPEG compression detection techniques identify whether or not an JPEG image file has undergone the compression twice, by knowing its embedded quantization table. This paper addresses another forensic scenario in which the quantization table of a JPEG file is not explicitly or reliably known, which may compel the forensic analyst to blindly reveal the recompression clues. To do this, we first statistically analyze the theory behind quantized alternating current (AC) modes in JPEG compression and show that the number of quantized AC modes required to detect double compression is a function of both the image’s block texture and the compression’s quality level in a fresh formulation. Consequently, a new double compression detection algorithm is proposed that exploits footprints introduced by all non-zero and zero AC modes based on Benford’s law in a low-dimensional representation via PCA. Then, some evaluation frameworks are constructed to assess the robustness and generalization of the proposed method on various textured images belonging to three standard databases as well as different compression quality level settings. The average F1-measure score on all tested databases in the proposed method is about 74 % much better than the state-of-the-art performance of 67.7 %. The proposed algorithm is also applicable to detect double compression from a JPEG file and localize tampered regions in actual image forgery scenarios. An implementation of our algorithms and used databases are available upon request to fellow researchers
  6. Keywords:
  7. Texture ; Algorithms ; Database systems ; Electric impedance measurement ; Image compression ; Vector quantization ; Double JPEG compressions ; Image forgery detections ; Quality levels ; Quantized AC modes ; Sparse signals ; Image texture
  8. Source: Journal of Mathematical Imaging and Vision ; Volume 54, Issue 3 , 2016 , Pages 269-286 ; 09249907 (ISSN)
  9. URL: https://link.springer.com/article/10.1007/s10851-015-0602-z