Biometric watermarking schemes based on QR decomposition and Schur decomposition in the RIDWT domain


Signal, Image and Video Processing, vol.18, no.3, pp.2783-2798, 2024 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 18 Issue: 3
  • Publication Date: 2024
  • Doi Number: 10.1007/s11760-023-02949-6
  • Journal Name: Signal, Image and Video Processing
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, zbMATH
  • Page Numbers: pp.2783-2798
  • Keywords: Biometric watermarking, Copyright protection, Firefly algorithm, QR decomposition, Redistributed invariant discrete wavelet transform, Schur decomposition
  • Karadeniz Technical University Affiliated: Yes


This paper introduces two robust and optimized watermarking techniques designed to protect the copyright of color digital images. These techniques involve embedding a biometric-based digital watermark, generated from an iris image, within the redistributed invariant discrete wavelet transform (RIDWT) domain. In the proposed algorithms, the Y channel of the color image in the YCbCr color space is initially decomposed using RIDWT. Then, the lowest frequency sub-band is partitioned into non-overlapping blocks. Blocks with lower standard deviations are selected to embed each bit of the watermark derived from the iris image. To conceal these watermark bits, two distinct approaches are used. In the first approach, selected blocks undergo a transformation using QR decomposition, while the second approach utilizes Schur decomposition for the same purpose. Before the embedding step, we optimize the embedding strength of the watermarking using the firefly algorithm, striking a balance between robustness and perceptual transparency. Experimental evaluations demonstrate that the proposed algorithms not only deliver good perceptual quality but also effectively resist various geometric and common image processing attacks. Our results underscore the superior robustness of these techniques compared to existing studies. Additionally, both methods maintain an exceptionally low error rate during ownership verification.