For an infrared hotbox monitoring system, when its visible image and infrared image have a large difference in resolution, the fusion of the images is not ideal. This is unfavorable for the determination of the hotboxes of a train. To solve this problem, an algorithm for fusing a visible image with a pseudo-color infrared image is proposed. In the algorithm, the gray levels of the pixels in the same part in both a visible image and an infrared image are reproduced firstly and bilinear interpolation is implemented for other parts in the images, so as to match the pixels in the visible image with those in the infrared image. Then, the infrared image is mapped by pseudo-dolor coding, and weighted fusion is implemented for the visible and infrared images in three channels (RGB). The experimental result shows that the fused image is richer in color. The location of the infrared target in the image is more accurate and is more convenient for the manual identification of the position of a hot-axis. The algorithm can meet the need for fusing infrared and visible images in an infrared hotbox monitoring system.