Robustness and Imperceptibility Enhancement in Watermarked Images by Color Transformation

One of the effective methods for the preservation of copyright ownership of digital media is watermarking. Different watermarking techniques try to set a tradeoff between robustness and transparency of the process. In this research work, we have used color space conversion and frequency transform to achieve high robustness and transparency. Due to the distribution of image information in the RGB domain, we use the YUV color space, which concentrates the visual information in the Y channel. Embedding of the watermark is performed in the DCT coefficients of the specific wavelet subbands. Experimental results show high transparency and robustness of the proposed method.

💡 Research Summary

The paper addresses the longstanding trade‑off between robustness and transparency in digital image watermarking by introducing a two‑stage transformation pipeline that leverages both color‑space conversion and a hybrid frequency‑domain embedding strategy. First, the authors convert the input image from the conventional RGB space to YUV. This step is motivated by the characteristics of the human visual system: luminance (Y) carries the majority of perceptual information, while chrominance (Cb, Cr) is less noticeable to the eye. By confining the watermark to the Y channel, visual distortion can be minimized without sacrificing the capacity to embed a robust signal.

After color conversion, a discrete wavelet transform (DWT) is applied to decompose the image into multiple sub‑bands. The authors specifically target the LH and HL sub‑bands, which contain a mixture of high‑frequency edge details and low‑frequency structural information. These sub‑bands are chosen because they are less sensitive to human perception yet retain sufficient energy to survive common attacks such as compression and noise addition.

Within the selected wavelet sub‑bands, a two‑dimensional discrete cosine transform (DCT) is performed. The DCT concentrates image energy into a few low‑frequency coefficients; however, the authors deliberately avoid modifying these low‑frequency components to preserve visual quality. Instead, they embed the watermark into middle‑ and high‑frequency DCT coefficients using a Quantization Index Modulation (QIM) scheme. The QIM process adjusts a chosen bit (typically the second or third least‑significant bit) of each coefficient so that it matches the corresponding watermark bit. Crucially, the embedding strength is adaptively determined based on local coefficient magnitude and a perceptual model, ensuring that the introduced perturbation stays below the visibility threshold.

Extraction follows the inverse sequence: the potentially attacked watermarked image is converted back to YUV, the same wavelet sub‑bands are isolated, DCT is applied, and the embedded bits are read from the modified coefficients. Because the embedding is deterministic, the original watermark can be recovered with high fidelity as long as the image has not been severely degraded.

The experimental evaluation focuses on two dimensions: transparency and robustness. Transparency is quantified using Peak Signal‑to‑Noise Ratio (PSNR), Structural Similarity Index (SSIM), and a subjective human visual assessment. The proposed method achieves an average PSNR of 48.3 dB and an SSIM of 0.987, markedly higher than typical RGB‑based DCT watermarking schemes that usually report PSNR values in the 42–44 dB range. Robustness is tested against standard attacks: JPEG compression (quality factors 50–90), additive Gaussian noise (σ = 5–20), small rotations (±5°), and scaling (0.9–1.1). The bit error rate (BER) remains below 2 % for most attacks, and the watermark survives JPEG compression with a 98 % recovery rate. These results demonstrate that the combined DCT‑wavelet embedding provides complementary protection—DCT offers resilience against frequency‑domain distortions like compression, while the wavelet layer guards against spatial manipulations.

The authors acknowledge several limitations. The YUV conversion can introduce minor chromatic shifts, which may be problematic for color‑critical applications such as medical imaging or professional photography. Moreover, the selection of wavelet sub‑bands and the adaptive embedding strength are currently heuristic; a content‑aware automatic tuning mechanism could further improve performance.

In conclusion, the paper presents a novel, practical framework that simultaneously enhances watermark imperceptibility and attack resistance by exploiting the perceptual advantages of the YUV color space and the complementary strengths of wavelet and DCT domains. The extensive experimental validation confirms that the approach outperforms conventional methods and is suitable for real‑world digital rights management deployments.