Abstract: Electromagnetic pulse thermography (EPT) imaging, with its unique mechanism, suffers from low contrast and blurry defect details. Existing infrared image enhancement algorithms are ineffective for tiny weld defects. This paper presents an infrared image enhancement method combining Butterworth high-pass filtering with a gravity model based bilateral filtering improvement. Firstly, the original infrared image is Fourier-transformed into the frequency domain. Multiple high-pass filtering with different cut-off frequencies eliminate low-frequency noise, generating a multi-band detail layer. Then, guided filtering using this detail layer creates a base layer. Next, the two layers are weighted and fused. Finally, the fused image undergoes the improved bilateral filtering based on the gravity model to produce the enhanced image. A comparative analysis is conducted with other algorithms, including adaptive histogram equalization with brightness correction, image enhancement using an image enhancement model based on gravitational force and lateral inhibition networks, and the 3C enhancement algorithm. Experimental results show that the proposed method enhances weld defects while effectively reducing background noise. Compared with the other three algorithms mentioned in this paper, the proposed algorithm respectively increases the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) of the image by 25% and 23%, and achieves better visual effects.