Skip to content
Conference

RAW domain HDR low-light image reconstruction via joint optimization of neural exposure field, low-rank denoising, wavelet transform and digital watermarking

Oct 2026 · International Conference on Advanced Image Processing Technology · Vol 14354, pp. 143540B - 143540B-8 · 0 citations · 12 references
Engineering

Abstract

We improve the RAW domain high dynamic range (HDR) low-light image reconstruction algorithm based on the joint optimization of neural exposure field, low-rank denoising, discrete wavelet transforms (DWT) and robust digital watermarking. We construct a sensor-specific RAW domain degradation model that improves the characterization of noise, underexposure, and dynamic range compression in real low-light scenes, and avoids information loss caused by ISP processing. Using the neural exposure field module, we build a model with variable exposure features in image space to achieve adaptive exposure adjustment and HDR expansion. We introduce a low-rank denoising module integrating non-local similarities, which uses the low-rank structure and redundant information of image blocks to preserve edge texture details and suppress Gaussian mixture noise. We further embed discrete wavelet transform to decompose images into multi-scale sub bands, enabling detail-preserving denoising and exposure enhancement in high-frequency components while stabilizing low-frequency structural information. To protect image copyright and authenticity in professional photography, medical imaging, and surveillance scenarios, we add a robust digital watermarking module that embeds imperceptible copyright information into the RAW domain HDR reconstructed image without degrading visual quality and quantitative metrics. We jointly optimize the loss function to realize end-to-end collaborative optimization of exposure adjustment, denoising, wavelet multi-scale processing and watermark embedding. Experiments on public datasets (SID, RAW-LOL) show that our algorithm achieves a PSNR of 33.12 dB and an SSIM of 0.950, with optimized quantitative metrics and visual effects. It effectively alleviates overexposure and underexposure in bright and dark areas, provides copyright protection, and is suitable for high-quality low-light imaging in professional photography and medical imaging.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.