Aiming at the problems of spectral information missing and target edge blurring in current infrared and low light level image fusion algorithms, a target difference attention algorithm and Transformer fusion algorithm for infrared and low light level image fusion are proposed. Firstly, a low-light level reconstruction network is constructed by using residual structure, and the perception loss is constructed by using VGG-16 to preserve the background texture and brightness information in the low-light level image to the maximum extent. Then, the feature extraction network is constructed by combining CNN and Transformer to extract the complete features of the image. At the same time, in the target differential attention module, the difference operation and feature extraction are carried out on the infrared image and low-light image, and the obtained infrared differential image is enhanced by the channel attention mechanism. Then the output feature map of CNN feature extraction network is added element by element to enhance the infrared target feature. Then, the high frequency and low frequency information of features are captured by gradient retention module to improve the retention of texture details. Finally, the feature reconstruction network is used to reconstruct the fused image. The experimental results show that the fusion results are not only more consistent with the human visual system, but also the objective evaluation indexes of MI and VIF are increased by 44.6% and 21.2%, respectively, compared with other fusion methods.