A novel low complexity retinex-based algorithm for enhancing low-light images

Retinex-based algorithms, drawing inspiration from the human visual system biology, have emerged as favored techniques in literature for enhancing low-light images. These algorithms aim to mitigate the adverse effects of poor illumination conditions, such as- narrow gray range, low brightness, low contrast, color distortion, and noise- thereby rendering the images more suitable for both human observation and computer processing. This paper presents a low-complexity Improved Retinex-based Multi-Phase algorithm (IRBMP) designed specifically for low light image enhancement. Performance of the proposed algorithm is analysed on the benchmark low-light image dataset Ex-Dark in comparison to various state-of-the-art retinex-based as well as traditional algorithms- MSRCP, NPE, SRIE, RBMP, AHE, log-transform, gamma-transform, and adaptive-sigmoid-transfer-function(ASTF). The proposed algorithm outperforms the baseline RBMP as well as the comparison algorithms in both subjective and objective metrics, such as BRISQUE and NIQE, indicating improved image quality. Additionally, the proposed method demonstrates faster computational time in comparison to other Retinex-based approaches, making it a promising candidate for real-time image processing applications.