An Improved Dark Channel Prior for Image Dehazing and Enhancement of Underground Mine Images Under Dust-Fog and Complex Illumination

To address image degradation caused by dust/fog scattering and non-uniform illumination in mine surveillance footage, this paper proposes an improved dark channel prior method based on the HSV brightness component. The method constructs a dark channel using the V component of the HSV color space via morphological erosion, estimates atmospheric light through random sampling and lower-bound constraints, and refines and fuses the transmission map using multi-scale guided filtering; Gamma correction is applied after recovering the V component to enhance brightness in dark areas. Experimental images were obtained from the Gaojialiang Coal Mine; scenes with varying degrees of degradation were selected for comparative evaluation, and multiple YOLO models were used to assess the impact of the enhancement process on personnel detection. Averaged across six representative mine scenes, the proposed method achieved a mean gradient of 8.606, a mean edge density of 14.072%, and a mean NIQE of 3.279. Compared to the original images, edge density increased from 10.551% to 14.072%, and the NIQE score dropped from 4.175 to 3.279. Personnel detection case studies demonstrated that the enhancement process improved detection results in dusty and foggy scenes. The results indicate that the proposed method enhances the visibility of object contours and local details in the tested mine scenes, achieves favorable no-reference image quality assessment scores, and reduces computational complexity, thereby offering a valuable reference for image processing in mine visual surveillance.