针对煤矿井下采掘作业中采煤和除尘活动引发尘雾分布不均及复杂光照条件，导致视频图像模糊不清、信息量和细节丢失等问题，提出了一种井下采掘工作场景非均质图像去雾与增强技术。

首先对雾图进行区域分割，计算不同亮度区域的全局暗通道环境光均值，并与通过自适应伽马矫正和多尺度高斯滤波得到的局部亮通道环境光进行加权融合，以获得精确的环境光估计。为了保证图像细节的同时实现自然去雾效果，采用多尺度融合矫正技术处理透射图，并利用联合双边滤波得到精细化的透射图，结合大气散射模型，实现尘雾图像的清晰化。针对去雾后的图像整体较暗且对比度不足，进一步采用修正白平衡处理，将图像转换到HSV空间，提出自适应饱和度矫正和改进对比度增强算法，并结合拉普拉斯锐化提升图像的细节和对比度。

通过选取DCP、MRP、OSFD、MF-LIME、CEEF 5种算法处理真实典型的场景图像，并采用多项指标与本研究算法处理结果进行对比实验，结果表明：与新颖优秀算法的最优指标对比，提出算法相比CEEF在平均梯度的平均提升约为两倍，提升了图像的清晰度；相比MRP的信息熵平均降低约为1%，保留了更多图像信息；相比OSFD的标准差平均提升约为6%，改善了图像对比度；相比CEEF的FADE平均降低约为23%，能更有效地降低尘雾密度且运行速度较快，表现出更优越的性能。提出的算法能够有效提高煤矿井下采掘工作场景中模糊图像的视觉效果和图像质量，增强了其在工程应用中的实用性。

In coal mines, the uneven distribution of dust haze and complex illumination conditions caused by underground coal mining and dust removal lead to blurred video images, as well as the loss of information and details. Hence, this study proposed a dehazing and enhancement algorithm for heterogeneous images of underground mining environments.

Initially, hazy images were segmented into zones with different brightness values, for which the average ambient light intensity of global dark channels was calculated. The calculation results were integrated through weighting with the ambient light of local bright channels, which was obtained using adaptive gamma correction and multiscale Gaussian filtering. Consequently, accurate ambient light intensity estimates were determined. To preserve image details while achieving natural dehazing effects, transmission maps were processed using multiscale fusion correction technology and were then refined using joint bilateral filtering. Afterward, clear hazy images were obtained using the atmospheric scattering model. To further enhance the overall brightness and contrast of the dehazed images, white balance correction was performed. Specifically, images were converted into the hue-saturation-value (HSV) color space. Then, the details and contrast of images were enhanced using the proposed adaptive saturation correction and improved contrast enhancement algorithm, as well as Laplacian sharpening.

Images of typical, actual scenarios were processed using five algorithms: dark channel prior (DCP), maximum reflectance prior (MRP), optimal-scale fusion-based dehazing (OSFD), multiscale fusion – low light image enhancement (MF-LIME), and contrast enhancement and exposure fusion (CEEF). The processing results of these algorithms were those of the proposed algorithm based on multiple indicators. The results indicate that compared to the above novel and excellent algorithms in terms of their optimal indicators, the proposed algorithm exhibited that: (1) The average gradients were approximately twice those obtained by CEEF, suggesting elevated image clarity. (2) The average information entropy decreased by approximately 1% compared to that of MRP, implying more information preserved. (3) The standard deviation increased by approximately 6% on average compared to OSFD, representing improved image contrast. (4) The average fog aware density evaluator (FADE) value by approximately 23% compared to CEEF, implying an effective reduction in the haze Therefore, the proposed algorithm can effectively improve the visual effects and quality of blurred images of underground mining environments in coal mines, exhibiting high utility in engineering.