The paper introduces a novel image restoration method called Histogram Transformer (Histformer) to address adverse weather conditions such as rain, fog, and snow. The method is designed to unify the restoration of multiple weather degradations using a single model. Key contributions include: 1. **Dynamic-Range Histogram Self-Attention (DHSA)**: This module segments spatial features into bins based on intensity and applies self-attention within and between bins, allowing for the selective focus on dynamic-range spatial features. It enhances the model's ability to capture long-range spatial features. 2. **Dual-Scale Gated Feed-Forward (DGFF)**: This module integrates multi-range and multi-scale depth-wise convolutions to extract multi-range information, improving the model's ability to model visual characteristics effectively. 3. **Correlation Loss**: This loss function leverages the Pearson correlation coefficient to enforce the linear relationship between restored and ground-truth images, ensuring that the restored image follows the original intensity ranking. The method is evaluated on various datasets and compared with state-of-the-art approaches, demonstrating superior performance in terms of PSNR and SSIM metrics. The authors also conduct ablation studies to validate the effectiveness of each component and provide real-world application examples to showcase the practical applicability of the method.The paper introduces a novel image restoration method called Histogram Transformer (Histformer) to address adverse weather conditions such as rain, fog, and snow. The method is designed to unify the restoration of multiple weather degradations using a single model. Key contributions include: 1. **Dynamic-Range Histogram Self-Attention (DHSA)**: This module segments spatial features into bins based on intensity and applies self-attention within and between bins, allowing for the selective focus on dynamic-range spatial features. It enhances the model's ability to capture long-range spatial features. 2. **Dual-Scale Gated Feed-Forward (DGFF)**: This module integrates multi-range and multi-scale depth-wise convolutions to extract multi-range information, improving the model's ability to model visual characteristics effectively. 3. **Correlation Loss**: This loss function leverages the Pearson correlation coefficient to enforce the linear relationship between restored and ground-truth images, ensuring that the restored image follows the original intensity ranking. The method is evaluated on various datasets and compared with state-of-the-art approaches, demonstrating superior performance in terms of PSNR and SSIM metrics. The authors also conduct ablation studies to validate the effectiveness of each component and provide real-world application examples to showcase the practical applicability of the method.