Abstract
High dynamic range (HDR) images are becoming pervasive due to capturing or rendering of a wider range of luminance, but their special display equipment is difficult to be popularized because of high cost and technological problem. Thus, HDR images must be adapted to the conventional display devices by applying tone mapping (TM) operation, which puts forward higher requirements for intellectual property protection of HDR images. As the robustness presents regional diversity in the low dynamic range (LDR) watermarked image after TM, which is different from the traditional watermarking technologies, a concept of watermarking activity is defined and used to distinguish the essential distinction of watermarking between LDR image and HDR image in this paper. Then, a novel robust HDR image watermarking algorithm is proposed against TM operations. Firstly, based on the hybrid processing of redundant discrete wavelet transform and singular value decomposition, the watermark is embedded by modifying the structure information of the HDR image. Distinguished from LDR image watermarking, the high embedding strength can cause more obvious distortion in the high brightness regions of HDR image than the low brightness regions. Thus, a perceptual brightness mask with low complexity is designed to improve the imperceptibility further. Experimental results show that the proposed algorithm is robust to the existing TM operations, with taking into account the imperceptibility and embedded capacity, which is superior to the current state-of-art HDR image watermarking algorithms.
Introduction
High Dynamic Range (HDR) images capture a massive range of luminance—up to 10 orders of magnitude—far exceeding what standard monitors can display. To view these images on conventional screens, they must undergo Tone Mapping (TM). Tone Mapping is a non-linear process that compresses the dynamic range, often amplifying details in dark areas while squashing them in bright areas. For traditional digital watermarking, TM acts like a severe "attack" that destroys the embedded data. Most existing watermarking techniques are designed for Low Dynamic Range (LDR) images and fail when transplanted to the HDR domain because they don't account for how TM redistributes information across different brightness regions.
The Proposed Watermarking Framework
The authors propose a hybrid approach that moves away from simple pixel-level embedding to modifying the structural information of the image. The process begins by converting the HDR image to the HSV (Hue, Saturation, Value) domain, performing watermarking only on the V (luminance) channel to avoid color distortion.
Watermark Embedding
The embedding process utilizes Redundant Discrete Wavelet Transform (RDWT) and Singular Value Decomposition (SVD). Unlike standard DWT, RDWT provides a shift-invariant representation with higher redundancy, which enhances the watermark's immunity to noise and geometric changes. The LL (low-frequency) sub-band from the RDWT is divided into 4x4 blocks, and SVD is applied to extract structural features. The watermark is embedded by modifying the relationship between specific elements in the $U$ matrix.
A key innovation is the Perceptual Brightness Mask (PBM). In HDR images, high-brightness regions are surprisingly sensitive to watermarking distortion. The PBM adaptively selects which blocks are "safe" to watermark based on a logarithmic luminance threshold, ensuring the watermark remains invisible while maximizing capacity.
Watermark Extraction
The extraction process mirrors the embedding steps. The stego image is converted to the HSV domain, and the V channel undergoes RDWT. The same secret key (Perceptual Mask) used during embedding is required to identify the blocks containing the watermark.
Key Insights: Watermarking Activity
The authors define a new qualitative scalar called "Watermarking Activity" (WA) to measure how robust a watermark remains in different regions of an image after it has been tone-mapped. Their core finding is that a watermark's survival depends on its placement: in low-brightness regions, TM often amplifies the signal, making extraction easier; in high-brightness regions, TM compresses the signal, often leading to data loss. By embedding in the $U$ matrix (representing structure) rather than the $S$ matrix (representing luminance values), the watermark becomes resilient to the non-linear "squashing" of values inherent in Tone Mapping.
Experimental Results
The algorithm was evaluated using 30 HDR images and 26 different Tone Mapping Operators (TMOs). The results demonstrate significant improvements over state-of-the-art methods:
- Capacity: Achieved an embedding capacity of $10^4$ to $10^5$ bits, roughly 20 times higher than previous robust HDR methods.
- Robustness: The Bit Error Rate (BER) averaged around 12.42% across 26 TMOs, outperforming competing methods that often saw BERs exceeding 20-30%.
- Imperceptibility: High quality scores (VDP_Q between 84-100) were achieved using the HDR-VDP-2 metric, indicating the watermark is virtually undetectable.
Conclusion
This paper introduces a robust watermarking algorithm specifically designed for the unique challenges of HDR imaging and Tone Mapping. While the Perceptual Brightness Mask is computationally efficient, future work will focus on improving segmentation accuracy for complex textures and extending the technique to HDR video content.
Summary updated