Detecting multiple H.264/AVC compressions with the same quantisation parameters

Zhenzhen Zhang, Jianjun Hou, Yu Zhang, Jingyu Ye, Yunqing Shi

Research output: Contribution to journalArticle

7 Scopus citations

Abstract

Multiple-compression detection is of particular importance in video forensics, as it reveals possible manipulations to the content. However, methods for detecting multiple compressions with same quantisation parameters (QPs) are rarely reported. To deal with this issue, a novel method is presented in this study to detect multiple H.264/advanced video coding compressions with the same QPs. First, a new set, named ratio difference set (RDS), is proposed, which is calculated by identifying the quantised DCT coefficients whose values will be changed after re-compression. Then, a discriminative and fixed statistical feature set extracted from RDS of each video is obtained to serve as input for classification. With the aid of support vector machines, the extracted feature set is used to classify the videos that have undergone H.264 compressions twice or more from those compressed just once. Experimental results show that high classification accuracy and robustness against copy-move attack and frame-deletion attack can be achieved with the authors' proposed method.

Original languageEnglish (US)
Pages (from-to)152-158
Number of pages7
JournalIET Information Security
Volume11
Issue number3
DOIs
StatePublished - May 1 2017

All Science Journal Classification (ASJC) codes

  • Software
  • Information Systems
  • Computer Networks and Communications

Fingerprint Dive into the research topics of 'Detecting multiple H.264/AVC compressions with the same quantisation parameters'. Together they form a unique fingerprint.

  • Cite this