Current CNN-based methods for infrared and visible image fusion are limited by the low discrimination of extracted structural features, the adoption of uniform loss functions, and the lack of inter-modal feature interaction, which make it difficult to obtain optimal fusion results. To alleviate the above problems, a framework for multimodal feature learning fusion using a cross-attention Transformer is proposed. To extract rich structural features at different scales, residual U-Nets with mixed receptive fields are adopted to capture salient object information at various granularities. Then, a hybrid attention fusion strategy is employed to integrate the complementing information from the input images. Finally, adaptive loss functions are designed to achieve optimal fusion results for different modal features. The fusion framework proposed in this study is thoroughly evaluated using the TNO, FLIR, and LLVIP datasets, encompassing diverse scenes and varying illumination conditions. In the comparative experiments, HATF achieved competitive results on three datasets, with EN, SD, MI, and SSIM metrics reaching the best performance on the TNO dataset, surpassing the second-best method by 2.3%, 18.8%, 4.2%, and 2.2%, respectively. These results validate the effectiveness of the proposed method in terms of both robustness and image fusion quality compared to several popular methods.

How to recover geometric transformations is one of the most challenging issues in image registration. To alleviate the effect of large geometric distortion in multimodal remote sensing image registration, a scale and rotate transform prediction net is proposed in this paper. First, to reduce the scale between the reference and sensed images, the image scale regression module is constructed via CNN feature extraction and FFT correlation, and the scale of sensed image can be recovered roughly. Second, the rotation estimate module is developed for predicting the rotation angles between the reference and the scale-recovered images. Finally, to obtain the accurate registration results, LoFTR is employed to match the geometric-recovered images. The proposed registration network was evaluated on GoogleEarth, HRMS, VIS-NIR and UAV datasets with contrast differences and geometric distortions. The experimental results show that the number of correct matches of our model reached 74.6%, and the RMSE of the registration results achieved 1.236, which is superior to the related methods.