DR-DiT: Image Deraining using Diffusion Model with Transformer

Rain removal in images is a significant challenge in the field of computer vision, as it is necessary to restore the visual quality of images degraded by rain streaks or droplets. In this paper, we propose a novel approach for rain removal using the diffusion model with a Transformer-based structure instead of the traditional U-Net architecture. The Transformer architecture has recently gained popularity for its ability to capture long-range dependencies and attention mechanisms. To reduce the computational complexity of self-attention in the Transformer, we used multi-head external attention instead of multi-head self-attention. Additionally, we design a multi-scale gated feed-forward network for improved local information capture. Our method is trained on a large dataset of rainy and corresponding clean images, and its performance is evaluated on various test images with different rain densities. Our experimental results demonstrate that our method achieves excellent results in terms of both quantitative metrics and visual quality.