Uneven lighting images enhancement algorithm for coal mine underground based on multi-exposure fusion

Artificial lighting is commonly utilized for illumination in underground coal mines, but it often introduces challenges in image quality, such as overexposure near light sources and underexposure in distant regions. Additionally, the presence of floating coal dust particles and water droplet sprays exacerbates these issues by causing halos, referred to as glow noise, which further degrade the visual clarity of captured images. To address these challenges, this study presents a novel image enhancement method for underground coal mines, leveraging multi-exposure fusion techniques to mitigate problems associated with uneven illumination and complex noise interference. The proposed method is composed of three key components: generating high-exposure and low-exposure image sequences while extracting glow noise from degraded images; estimating initial weight maps for images at different exposure levels based on evaluation metrics; and refining the weight maps to reduce the impact of glow noise before fusing the image sequences into an enhanced output. The enhancement process begins with the use of a camera response model to boost the pixel exposure of the original image, producing a high-exposure image. Meanwhile, a glow model is employed to eliminate glow effects, resulting in a low-exposure image and enabling the extraction of glow noise. Following this, the original image is assessed for contrast, saturation, and exposure to compute initial weight maps for images at varying exposure levels. These weight maps are then optimized to suppress the influence of glow noise by reducing the pixel weights in regions affected by such noise, generating the final weight maps. Finally, the multi-exposure image sequences are integrated with the optimized weight maps through pyramid decomposition and reconstruction fusion, yielding an enhanced image. To evaluate the effectiveness of the proposed method, extensive comparative experiments were conducted. The results demonstrated that the method achieved superior overall performance compared to existing approaches, effectively addressing glow noise and uneven illumination issues in underground coal mine images. This approach significantly enhances the quality of monitoring images, providing critical support for coal mine safety operations and advancing the development of intelligent mining systems.