Retinex mine image enhancement algorithm based on TopHat weighted guided filtering

The uneven distribution of light sources and weak light in coal mines lead to low brightness and unclear image. The traditional Retinex algorithm has the problems of detail loss, edge blur and halo when processing low illumination images of coal mines. In order to solve the above problems, a new algorithm named THWGIF-Retinex based on TopHat weighted guided filtering is proposed to enhance the mine image. Firstly, the image is transformed from RGB space to HSV space. Then the image is separated into three channel components of hue, saturation and brightness. Secondly, the TopHat transform is used to improve the weight factor of the weighted guided filtering. The illumination component of the image is extracted from the brightness component. The edge enhancement of the brightness component is realized. Thirdly, the illumination component and the saturation component are corrected by adopting a self-adaptive gamma correction function. The reflection component is obtained from the illumination component by the Retinex algorithm. The details and color effect of the image light source are further improved. Finally, the hue component, the corrected saturation component and the reflection component are combined and converted to RGB space to obtain an enhanced mine image. The THWGIF-Retinex algorithm, multi-scale Retinex (MSR) algorithm and weighted guided filtering Retinex (WGIF-Retinex) algorithm are compared and verified from subjective evaluation and objective evaluation. The subjective evaluation results show that the original image of low illumination without strong light is enhanced by the THWGIF-Retinex algorithm. The color reproduction degree of the image is higher, the image edge is clearer, and the visual effect is obviously enhanced. The THWGIF-Retinex algorithm has a good effect on halo reduction for the mine low-illumination original image with strong light. The THWGIF-Retinex algorithm is better than the WGIF-Retinex algorithm in restoring the details and clarity of dark areas. The objective evaluation results show that the information entropy, the average gradient, the standard deviation and the no-reference structural sharpness (NRSS) of the image enhanced by the THWGIF-Retinex algorithm are increased by 12.50%, 109.07%, 52.44% and 45.46% respectively for the low illumination images without strong light. Compared with the MSR algorithm, the information entropy, average gradient, standard deviation and NRSS of the image enhanced by the THWGIF-Retinex algorithm are increased by 1.24%, 81.44%, 18.23% and 36.67% respectively for the mine low illumination image with strong light. Compared with the WGIF-Retinex algorithm, the THWGIF-Retinex algorithm has lower information entropy. However, the average gradient and NRSS are improved by 72.34% and 23.87% respectively.