2D Image Watermarking
Explanation: Digital watermarking is a technology to conceal information in audio, image, and vide for the purpose of content authentication, authenticity identification, distribution tracking, copyright protection, and so on. The inserted watermark should be invisible to the human eye and be robust enough to be detected even if various attacks are applied. For the security of 2D images, we provide a novel 2D image watermarking method that is robust against various attacks and enhances security vulnerability of the technology itself unlike existing methods. You can insert a watermark in the image by setting a number between 0 and 65535 as the password, and verify whether the watermark is inserted in the image using the entered password.
Embedding
Select the image to be watermarked
(Filename must be written in English or numeric)

Set its password
Detection
Select the watermarked image to be verified
(Filename must be written in English or numeric)

Insert its password


Set JPEG attack strength

Set Noise attack strength

Set filtering attack strength
Technology description
  • Attack strength specification
    • JPEG attack strength
      • Weakly = JPEG quality 90
      • Moderately = JPEG quality 80
      • Hardly = JPEG quality 70
    • Noise attack strength
      • Weakly = Gaussian noise~N(6,1)
      • Moderately = Gaussian noise~N(9,1)
      • Hardly = Gaussian noise~N(12,1)
    • Filtering attack strength
      • Weakly = Gaussian lowpass filter, kernel size=5x5, standard deviation=0.5
      • Moderately = Gaussian lowpass filter, kernel size=5x5, standard deviation =0.65
      • Hardly = Gaussian lowpass filter, kernel size=7x7, standard deviation =0.8

  • Types of attacks covered (NC is normalized correlation, PSNR=45dB)
    • Cropping (factor=0.8), NC=0.7188
    • Rescaling (factor=0.6, NC=1.0000
    • Translation (shift=2), NC=0.9208
    • Rotation (1°, 10°), NC=0.9646, 0.8396
    • General affine transform, NC=1.0000
    • Row and column removal, NC=0.9182
    • Gamma correction (γ=0.3), NC=1.0000
    • Gaussian filtering (σ=1), NC=1.0000
    • Median filtering (kernel size=3x3), NC=0.9958
    • JPEG (factor=80), NC=0.4250
    • Gaussian noise addition (σ=0.05), NC=0.9958
    • Salt-and-pepper noise addition (p=0.01) = 1.0000