USwin-HDRI: High Dynamic Range Imaging in Dynamic Scenes Combining Unet and Transformer

High Dynamic Range Imaging (HDRI) is crucial for capturing scenes with a wide range of luminance levels. Traditional methods often struggle with dynamic scenes due to issues like camera shake and object motion, leading to artifacts in the resultant images. To address these challenges, this paper presents a novel method for High Dynamic Range Imaging (HDRI) in dynamic scenes, called USwin-HDRI, which integrates UNet and Transformer technologies to address imaging challenges. The proposed network architecture includes an encoder, feature alignment subnet, feature enhancement subnet, decoder, and loss function design and implementation. By utilizing gamma correction, deformable convolutions, and the Swin Transformer, our method effectively aligns features, enhancing image consistency and detail preservation. Experimental results demonstrate that USwin-HDRI excels in reducing ghosting and maintaining image clarity compared to several traditional exposure fusion algorithms, showing superior performance in metrics such as PSNR and SSIM. Both subjective and objective analyses confirm the effectiveness of this method.