A Novel Method for Block Size Forensics Based on Morphological Operations

  • Authors:

  • Weiqi Luo;Jiwu Huang;Guoping Qiu

  • Affiliations:

  • Guangdong Key Lab. of Information Security Technology, Sun Yat-Sen University, Guangdong, China 510275;Guangdong Key Lab. of Information Security Technology, Sun Yat-Sen University, Guangdong, China 510275;School of Computer Science, University of Nottingham, UK NG 8, 1BB

  • Venue:
  • Digital Watermarking
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Passive forensics analysis aims to find out how multimedia data is acquired and processed without relying on pre-embedded or pre-registered information. Since most existing compression schemes for digital images are based on block processing, one of the fundamental steps for subsequent forensics analysis is to detect the presence of block artifacts and estimate the block size for a given image. In this paper, we propose a novel method for blind block size estimation. A 2脳2 cross-differential filter is first applied to detect all possible block artifact boundaries, morphological operations are then used to remove the boundary effects caused by the edges of the actual image contents, and finally maximum-likelihood estimation (MLE) is employed to estimate the block size. The experimental results evaluated on over 1300 nature images show the effectiveness of our proposed method. Compared with existing gradient-based detection method, our method achieves over 39% accuracy improvement on average.