A Comparative Study of Classical Digital Watermarking Techniques for High Visual Quality

Abstract

Digital image watermarking has become a critical technology in copyright protection and ownership authentication. This paper presents a comparison study amongst four classical watermarking schemes: The LSB (Least Significant Bit), DCT (Discrete Cosine Transform), DWT (Discrete Wavelet Transform), and hybrid DWT-DCT methods. The evaluation will concern visual imperceptibility under no attack conditions by considering PSNR (Peak Signal to Noise Ratio) as the metric.A special feature of this work consists of using a binary QR code of 32×32 pixels carrying semantic information the author's name as the watermark, inserted into the standard Lena image (512×512 and 256×256 pixels, 24-bit RGB). This practical choice allows direct identification by a smartphone and makes this method more applicable in real applications.Experimental results depict that all the methods yield PSNR values greater than 50 dB, which gives excellent quality and imperceptibility. The LSB method achieved the highest PSNR equal to 75.36 dB for 512×512 and outperformed the existing literature; it can thus be said that the transparency in such methods is almost perfect. DWT and DCT have PSNR values of 51.27 dB and 51.14 dB, respectively, while the hybrid DWT-DCT has a PSNR of 50.57dB. While the LSB is excellent in terms of invisibility, it is inherently fragile and can easily be destroyed by host image processing different attacks such JPEG compression, median filtering, random cropping, and additive Gaussian noiseIn contrast, the transform domain methods DCT, DWT, and DWT-DCT embed the watermark in more stable frequency coefficients and thus provide a better tradeoff between imperceptibility and robustness.This work addresses a critical lacuna in the literature by offering a direct comparison of classical methods on standard images using a semantically rich QR code. It paves the way for future research on security enhancement, robustness evaluation under various attacks, adaptive embedding strategies, and blind extraction mechanisms.