Brightness Perceiving for Recursive Low-Light Image Enhancement

TL;DR

This work tackles low-light enhancement under high dynamic range by framing enhancement as a recursive, brightness-aware process. It introduces ACT-Net for adaptive contrast/texture enhancement and BP-Net to predict recursion depth from brightness distribution, all trained with an unsupervised strategy. The approach achieves state-of-the-art results across multiple metrics on mixed brightness datasets and real-world data, while maintaining favorable computational efficiency. The framework generalizes to multi-exposure correction and points to future work including denoising and expanded real-world data for robustness.

Abstract

Due to the wide dynamic range in real low-light scenes, there will be large differences in the degree of contrast degradation and detail blurring of captured images, making it difficult for existing end-to-end methods to enhance low-light images to normal exposure. To address the above issue, we decompose low-light image enhancement into a recursive enhancement task and propose a brightness-perceiving-based recursive enhancement framework for high dynamic range low-light image enhancement. Specifically, our recursive enhancement framework consists of two parallel sub-networks: Adaptive Contrast and Texture enhancement network (ACT-Net) and Brightness Perception network (BP-Net). The ACT-Net is proposed to adaptively enhance image contrast and details under the guidance of the brightness adjustment branch and gradient adjustment branch, which are proposed to perceive the degradation degree of contrast and details in low-light images. To adaptively enhance images captured under different brightness levels, BP-Net is proposed to control the recursive enhancement times of ACT-Net by exploring the image brightness distribution properties. Finally, in order to coordinate ACT-Net and BP-Net, we design a novel unsupervised training strategy to facilitate the training procedure. To further validate the effectiveness of the proposed method, we construct a new dataset with a broader brightness distribution by mixing three low-light datasets. Compared with eleven existing representative methods, the proposed method achieves new SOTA performance on six reference and no reference metrics. Specifically, the proposed method improves the PSNR by 0.9 dB compared to the existing SOTA method.

Figures (12)

• Figure 1: (a, b) Comparison with existing methods on PSNR and SSIM under different brightness levels. (c) Visualization comparison with existing representative multi-exposure enhancement method FEC huang2022deep and low light enhancement method SCI ma2022toward in 4 different brightness of the same background.
• Figure 2: (a) The t-SNE van2008visualizing of brightness distribution of existing three prevailing low-light datasets. (b) The visualization of details degradation in different low-light scenes from Level_1 (darkest) to Level_4 (brightest). We can observe that the brightness distribution of low-light images varies significantly under different brightness levels; the lower the brightness, the greater the structural damage.
• Figure 3: The overall architecture of our recursive enhancement framework, which consists of the ACT-Net and BP-Net.
• Figure 4: Qualitative comparison with existing state-of-the-art methods. From the top to bottom row are the low-light conditions from Level_1 to Level_4, where Level_1 represents the darkest low-light images, and Level_4 represents the brightest.
• Figure 5: Comparison of enhancement results in the same scene with existing representative methods. From the left to right in each image are the low-light conditions from darkest to brightest.