Thermal fault is a common fault type in the power equipment. This paper introduces a thermal fault diagnosis method for the power equipment by employing guided filter. The proposed method consists of two main steps. First, according to the temperature difference between the thermal fault area and the background in infrared images, the Mahalanobis distance between the fault area and the background is estimated, and the initial thermal fault diagnosis results can be generated. Then, guided filter is used to improve the diagnosis performance, by utilizing the spatial correlation between adjacent pixels fully utilized. Experimental results demonstrate that the proposed diagnosis method has a better diagnosis performance than the current state-of-the-art detectors.