In order to solve the problem of real-time image fusion and the disparity between the infrared and low level light(LLL) image field,a window registration and fusion method for infrared and LLL images based on affine transformation is proposed. Firstly,we use LLL image with large field of view as background to open the interested target area,the size of the window is determined by the hardware speed and the amount of computation of the registration and fusion algorithm. Then we use bilinear interpolation and affine transformation to establish a corresponding matching relationship between the pixels of the infrared and LLL image to complete the fast window image registration and fusion. The results of window fusion were analyzed and evaluated. The experimental results show that this method not only reduces the image fusion processing data,but also preserves the important details fusion information. It can effectively improve the real-time performance of image fusion under the condition of satisfying the needs of human observation. So it has high application value for image fusion system with both hardware speed and real-time requirements.