关键词: 图像去噪　 , 伽马变换　, LIME算法　, 图像分解　, 矿井智能化

Abstract: Mine monitoring images suffer from low signal-to-noise ratio and high dynamic range noise due to low illumi nation, non-uniform artificial lighting, and dust or fog interference,which limits the accuracy of safety monitoring results.To enhance image clarity in mine monitoring,a novel image denoising algorithm combining gamma transformation with Local Inter pretable Model-agnostic Explanation (LIME) algorithm is proposed.Firstly,the LIME algorithm is applied to enhance illumi nation in raw mine images,while adaptively adjusting the gamma parameter based on the image mean to achieve pixel-wise brightness correction,thereby improving overall brightness and balancing light distribution,and preliminarily restoring details in dark regions.Next,an image decomposition technique based on the atmospheric scattering model separates the image into scene radiance,transmission,and global atmospheric light components.A spatially adaptive exposure adjustment is then implemented by constructing an exposure matrix for the transmission component using a camera response model,followed by linearly weigh ted fusion of the transmission component to remove dust and fog noise.Finally,the denoised and enhanced image is reconstruc ted using the Koschmieder model.The proposed method is compared experimentally with several state-of-the-art algorithms,in cluding improved Enlighten-GAN,multi-weight fused Retinex,three-stream three-channel color-balanced defogging,and Z DCE-DNet.Results show that the proposed method achieves a Gradient Magnitude Similarity Deviation (GMSD) of 0.359,an information entropy-weighted Structural Similarity Index (SSIM) of 0.947,and a multi-scale structural similarity index of 0.971,demonstrating significant performance advantages.This algorithm provides an effective solution for improving both effi ciency and quality in mine monitoring image processing and offers valuable reference for advancing intelligent image processing applications in underground mining environments.

Key words: imagedenoising,gammatransform,LIMEalgorithm,imagedecomposition,mineintelligentization