Abstract: Due to the influence of dust, humidity, and illumination, the underground monitoring video is prone to problems such as uneven lighting, low contrast, and blurred details, which affects the application of visual servo system and video analysis system. Traditional image enhancement algorithms tend to be excessive for bright area and insufficient for detail of dark area, which lead to unnatural halo and color distortion. Aiming at this problem, a mine image enhancement algorithm based on retinex using a multi-weight fusion strategy is proposed. In the HSV space, the algorithm keeps the hue unchanged, and only enhances the brightness and the saturation. Firstly, a multi-scale gradient domain guided filter is adopted to estimate the illuminance component from the brightness component, and an adaptive gamma function is applied to correct the illuminance considering the coexistence of overexposure and underexposure in mine image. At the same time, the contrast of the reflection component is enhanced by using an adaptive histogram equalization algorithm with limited contrast. After that, the brightness-corrected illuminance component and the contrast-enhanced reflection component are fused with multiple weights, in which the normalized weight is composed of local contrast, brightness and sharpness weight. Aiming at the problem of color migration, a new mapping function is proposed to conduct a nonlinear extension of the saturation component. Finally, the image is converted from the HSV space to the RGB space to complete the image enhancement. The experimental results show that the proposed algorithm has better performance than that of MSRCR, NPE, SRIE, and BIME algorithms in average gradient, information entropy and standard deviation. The algorithm improves the contrast, clarity and color accuracy, and suppresses haloing and artifacts.