Local multi-watermarking method based on robust and adaptive feature extraction

Xiao Chen Yuan, Mianjie Li

Research output: Contribution to journalArticlepeer-review

36 Citations (Scopus)


This paper proposes a local multi-watermarking method based on robust and adaptive feature extraction. The Robust and Adaptive Feature Detector based on DAISY Descriptor (RAF3D) is proposed to extract the feature regions of high robustness and stability. The multi-watermarking method is proposed to embed the multiple watermarks simultaneously into the same extracted feature region. In this way, the capacity will be flexible with either the number of feature regions or the number of watermarks. In the proposed method, the Gram–Schmidt process is applied to embed the watermarks in orthogonal spaces, which guarantees the multiple watermarks can be extracted independently. By repeatedly embedding the watermarks into the numerous feature regions, the success rate of watermark detection can be greatly strengthened. In addition, the local embedding strategy improves the imperceptibility of the watermarked image. Extensive experiments are conducted to evaluate the performance of the proposed scheme and the comparison with several existing methods demonstrate that the proposed scheme outperforms the existing methods in terms of the robustness against various attacks.

Original languageEnglish
Pages (from-to)103-117
Number of pages15
JournalSignal Processing
Publication statusPublished - Aug 2018
Externally publishedYes


  • Gram–Schmidt process
  • Local multi-watermarking
  • Robust and adaptive feature detector


Dive into the research topics of 'Local multi-watermarking method based on robust and adaptive feature extraction'. Together they form a unique fingerprint.

Cite this