This paper proposes a method of extracting fault regions from infrared images of electronic equipment using a pulse-coupled neural network (PCNN); the method is based on iterative clustering. The PCNN model is used as the kernel method for image processing. Several parameters are first determined, and then hierarchical clustering is introduced to enable the PCNN model to segment an image into multiple regions on the basis of the inner similarity. In addition, the cluster centers are computed and sorted from large to small to find the pulse region with the highest brightness. The merging processing is final carried out with its neighboring region and the measure of similarity. The method thus improves the ability of the PCNN model to segment infrared images efficiently and effectively identifies thermal fault regions. Experimental results show that the proposed method exhibits good segmentation performance and is suitable for processing infrared images of thermal fault regions.