Color Image Blind Watermarking Algorithm Based on QR Decomposition and Voting in DWT Domain

Junxiang Wang, Ying Liu, Yonghong Zhu
2016 International Journal of Security and Its Applications  
To achieve a better trade-off between robustness and imperceptibility, a blind watermarking algorithm based on QR decomposition and voting for color image in DWT domain is proposed in the paper. Most of the existing QR decomposition based watermarking schemes take the first row of R matrix into consideration while in our approach the watermark is embedded by modifying the first column element of Q matrix. In the embedding process, DCT is applied to the R, G and B color components, respectively.
more » ... Then DWT is utilized on the DCT coefficients of each component and acquire corresponding LL, HL, LH, and HH. Later, the associated DWT coefficients in LH and HH are decomposed with QR decomposition and the watermark message is embedded into the first column of Q matrix by changing Q 21 and Q 31 with stable characteristic. In the extraction phase, a redundant watermark scheme with voting method is used to improve the robustness of the watermarking algorithm. Experimental results show that the proposed algorithm, compared with the existing methods is robust enough to resist common signal attacks including filtering, noising, small cutting and JPEG compressing. 40 Copyright ⓒ 2016 SERSC Robustness Analysis To verify the robustness of the above-mentioned algorithm, we performed some common attack signal processing operations on the watermarked Lena image, Such as, cropping (Cr) including 1/4 corner-cropping (Ccr) and side-cropping (Scr), Scaling (Scl), median filtering (Md), Low-pass filtering (Lp) , mean filtering (Mf), Gaussian filter (Gf), Gaussian noise (Gn), Salt and pepper noise (Spn) and speckle noise (Sn). Some results indicated in Fig.6 demonstrated the effectiveness of the watermarking scheme proposed in this paper. Three independent comparative experiments have been done. First, we compare our scheme with the method proposed in [7] . Here, we adopt AR as evaluating indicator and the results are shown in Table. 3. In the second experiment, we compare NCC for the proposed scheme and method mentioned in [8] and the results are shown in Fig.7 and Tab.4. Final, we compare our voting scheme with non-voting method, the result in Fig.8 shows the NCC of voting method for some standard images, such as "peppers" and "baboon", is higher than the non-voting method.
doi:10.14257/ijsia.2016.10.11.04 fatcat:hkw45oq6yzbs3ifzwrerbsovry