Hadamard based watermarking using structural correlation average

Digital Watermarking provides good information security, especially for copyright protection. Meanwhile, the main problem of the watermarking system is a balance between imperceptibility and robustness to get optimum results. There are two main methods in watermarking, which is transformation and spatial method. Transform domain has a good performance but high computational cost, which is unsuitable for current needed in real-time computing. Meanwhile, most spatial methods are fast, especially integer-based ones like Hadamard. However, it is very difficult to determine the quantization size. This paper proposed a model for determining the best quantization value which can work adaptively. The value is determined using the structural correlation between host, watermark and watermarked image. The result shows that there are three optimum values, namely 38, 39, and 40. The best value is achieved in 39, which result in great imperceptibility with an SSIM value of 0.995. The robustness test of JPEG compression, Gaussian filtering, resizing, cropping, gamma correction, rotation, and contrast adjustment results in good NC values of 0.903, 0.933, 0.976, 0.972, 0.633, 0.601, and 0.890, respectively. This shows that the proposed model can define the proper quantization size, resulting in optimum imperceptibility and robustness.