AMDCENR: Attention Mechanism and Deep Curve Estimation under Noise Reduction Network for Low-Light Image Enhancement

Low-light photos can present difficulties, including reduced brightness, poor contrast, and noise, which makes them unsuitable for a variety of image processing uses. To address these problems, previous techniques for improving low-light photos, based on the zero-reference depth network model (Zero-DCE), have attempted to estimate both pixel-wise and higher-order curves via an iterative process. However, the intrinsic noise in low-light photos has not been taken into account by these methods up to this point, which has a substantial impact on the final image quality. To overcome this drawback, we suggest a brand-new method called Attention Mechanism and Deep Curve Estimation under Noise Reduction Network (AMDCENR-net), which is especially made for improving low-light photos. In order to improve the lightmap and allow picture denoising, we construct a Noise Reduction Module (NRM) that makes use of regularization models and elemental subtraction iterations. Moreover, our method is strengthened by combining both channel attention and spatial attention processes to improve picture quality and reduce information loss. Channel attention is used to reduce less important channels in the image and emphasize the more informative ones. Spatial attention, on the other hand, focuses on highlighting visually significant areas of the image. Extensive experiments validate the effectiveness of our suggested strategy, showing both qualitative and quantitative improvements over current state-of-the-art techniques. These results highlight the AMDCENR-net's effectiveness in handling the challenges related to low-light picture enhancement.