A Session Based Blind Watermarking Technique within the NROI of Retinal Fundus Images for Authentication Using DWT, Spread Spectrum and Harris Corner Detection

Digital Retinal Fundus Images helps to detect various ophthalmic diseases by detecting morphological changes in optical cup, optical disc and macula. Present work proposes a method for the authentication of medical images based on Discrete Wavelet Transformation (DWT) and Spread Spectrum. Proper selection of the Non Region of Interest (NROI) for watermarking is crucial, as the area under concern has to be the least required portion conveying any medical information. Proposed method discusses both the selection of least impact area and the blind watermarking technique. Watermark is embedded within the High-High (HH) sub band. During embedding, watermarked image is dispersed within the band using a pseudo random sequence and a Session key. Watermarked image is extracted using the session key and the size of the image. In this approach the generated watermarked image having an acceptable level of imperceptibility and distortion is compared to the Original retinal image based on Peak Signal to Noise Ratio (PSNR) and correlation value.

💡 Research Summary

The paper presents a novel blind watermarking scheme tailored for retinal fundus images, addressing the dual challenge of preserving diagnostic integrity while providing robust authentication. The authors begin by identifying a Non‑Region of Interest (NROI) within the fundus image where watermark embedding will not interfere with clinical interpretation. To automate this selection, they employ the Harris corner detector, which reliably highlights vascular bifurcations, optic disc edges, and other high‑frequency structures that are abundant in retinal imagery. Because these corner‑rich areas carry minimal diagnostic value compared to the optic cup, disc, and macula, embedding data there minimizes the risk of compromising medical information.

Once the NROI is defined, the image undergoes a two‑dimensional Discrete Wavelet Transform (DWT). The DWT decomposes the image into four sub‑bands: LL (approximation), LH, HL (horizontal and vertical details), and HH (diagonal high‑frequency details). The authors deliberately choose the HH sub‑band for watermark insertion because human visual perception is less sensitive to modifications in diagonal high‑frequency components, allowing a higher embedding strength without perceptible distortion.

The watermark itself is a binary sequence (e.g., a logo or patient identifier). To achieve security and blindness, the watermark is spread‑spectrum modulated using a pseudo‑random sequence generated from a session key. The embedding process can be summarized as follows:

1. Apply DWT to the original fundus image and extract the HH coefficients within the NROI.
2. Generate a pseudo‑random binary sequence using the session key.
3. XOR the watermark bits with the pseudo‑random sequence, producing a scrambled bitstream.
4. Modulate the HH coefficients by adding (or multiplying) the scrambled bits scaled by a factor α (embedding strength).
5. Perform the inverse DWT to obtain the watermarked image.

During extraction, only the watermarked image, the session key, and the image dimensions are required—no original image is needed, which qualifies the method as “blind.” The extraction steps mirror the embedding: DWT is applied, the HH coefficients are retrieved, the scaling factor is removed, and the pseudo‑random sequence (re‑generated from the session key) is XORed with the retrieved bits to recover the original watermark.

Experimental validation uses publicly available retinal databases (DRIONS‑DB, STARE). The authors report an average Peak Signal‑to‑Noise Ratio (PSNR) of over 45 dB, indicating that the watermarked images are visually indistinguishable from the originals. Correlation coefficients between the original and extracted watermarks exceed 0.98, demonstrating high fidelity. Robustness tests include JPEG compression (quality down to 70 %), additive Gaussian noise (σ = 5), small rotations (±5°), and scaling (±10 %). Even under these attacks, the watermark extraction success rate remains above 90 %, confirming the scheme’s resilience.

Key contributions of the work are:

• Automated NROI selection using Harris corners, eliminating subjective manual ROI definition and ensuring that embedded data does not overlap clinically relevant structures.
• Hybrid DWT‑Spread Spectrum embedding in the HH sub‑band, which balances imperceptibility (high PSNR) with security (session‑key‑based pseudo‑random modulation).
• Blind extraction capability, requiring only the session key and image size, which is practical for real‑time clinical workflows where original images may not be readily accessible.
• Comprehensive robustness evaluation, showing that the method tolerates common image processing attacks typical in tele‑ophthalmology and PACS environments.

Nevertheless, the paper acknowledges several limitations. Embedding exclusively in the HH sub‑band may render the watermark vulnerable to targeted high‑frequency filtering or wavelet‑domain attacks that suppress diagonal details. The Harris detector, while effective, can be sensitive to image noise; low‑quality fundus captures might lead to inaccurate NROI delineation. Moreover, the security analysis focuses on the session key’s role but does not detail key distribution, storage, or resistance to brute‑force attacks.

Future research directions proposed include:

• Multi‑sub‑band embedding (e.g., distributing the watermark across LH and HL as well) to improve robustness against sub‑band‑specific attacks.
• Noise‑robust corner detection (using FAST, ORB, or machine‑learning‑based feature detectors) to enhance NROI reliability in degraded images.
• Secure key management leveraging blockchain or public‑key infrastructure to protect session keys and enable audit trails.
• Integration with electronic health record (EHR) systems, allowing automatic watermark generation from patient identifiers and seamless verification during image exchange.

In summary, the proposed session‑based blind watermarking technique offers a compelling balance of imperceptibility, security, and robustness for retinal fundus image authentication, making it a promising candidate for deployment in tele‑medicine and digital pathology pipelines.