RGGB sensor arrays are commonly used in digital cameras and mobile photography. However, images of extreme dark-light conditions often suffer from insufficient exposure because the sensor receives insufficient light. The existing methods mainly employ U-Net variants, multi-stage camera parameter simulation, or image parameter processing to address this issue. However, those methods usually apply color adjustments evenly across the entire image, which may cause extensive blue or green noise artifacts, especially in images with dark backgrounds. This study attacks the problem by proposing a novel multi-step process for image enhancement. The pipeline starts with a self-attention U-Net for initial color restoration and applies a Color Correction Matrix (CCM). Thereafter, High Dynamic Range (HDR) image reconstruction techniques are utilized to improve exposure using various Camera Response Functions (CRFs). After removing under- and over-exposed frames, pseudo-HDR images are created through multi-frame fusion. Also, a comparative analysis is conducted based on a standard dataset, and the results show that the proposed approach performs better in creating well-exposed images and improves the Peak-Signal-to-Noise Ratio (PSNR) by 0.16 dB compared to the benchmark methods.

Currently, most images captured by high-pixel devices such as mobile phones, camcorders, and drones are in RGGB format. However, image quality in extremely dark scenes often needs improvement. Traditional methods for processing these dark RGGB images typically rely on end-to-end U-Net networks and their enhancement techniques, which require substantial resources and processing time. To tackle this issue, we first converted RGGB images into RGB three-channel images by subtracting the black level and applying linear interpolation. During the training stage, we leveraged the computational efficiency of the state-space model (SSM) and developed a Mamba U-Net end-to-end model to enhance the restoration of extremely dark RGGB images. We utilized the see-in-the-dark (SID) dataset for training, assessing the effectiveness of our approach. Experimental results indicate that our method significantly reduces resource consumption compared to existing single-step training and prior multi-step training techniques, while achieving improved peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) outcomes.