In geological exploration, high-resolution borehole images are critical to understanding the structure, composition, and properties of the underlying layer. However, due to limitations of imaging conditions during the borehole process, it is difficult to directly obtain high-resolution borehole images. To address this problem, this paper constructs a borehole image super-resolution model (BISRM) by combining a multi-region visual attention mechanism and a conditional diffusion model. At the same time, we introduce modulation coefficients to enhance the generalization ability of the model and improve the quality of super-resolution images. In addition, in order to verify the feasibility of the borehole image super-resolution model, we collected a total of 160,000 low-resolution borehole images from different geological exploration sites and constructed a low-resolution borehole image dataset Our BISRM was experimentally verified under three environments: sunlight, the coexistence of sunlight and cold light sources, and cold light sources. Experimental results show that our model is better than the comparison method in terms of IS (saliency score) and FID (generated image quality distribution) indicators. The effectiveness of BISRM is significative to fields such as geotechnical engineering safety analysis and intelligent geological exploration.