<p>Remote sensing images play an important role in Earth observation and practical applications. However, under complex weather conditions, they are easily affected by factors such as haze and raindrops, leading to significant degradation in image quality. To address the challenges of spatially non-uniform degradation and scale diversity in remote sensing image restoration under adverse weather conditions, we propose a dynamic hybrid Transformer–CNN network, called DHTCformer. The proposed method effectively integrates the contextual modeling capability of Transformers with the local feature extraction strength of CNNs, thereby improving the representation of weather-degraded images. Specifically, a dynamic dual-path Transformer module is designed by introducing parallel adaptive channel attention and dynamic context-aware spatial attention, which dynamically focuses on critical degraded regions from both channel and spatial domains, thereby effectively alleviating the impact of spatially non-uniform weather degradation in remote sensing images. In addition, a multi-scale gated feed-forward network is incorporated to achieve efficient cross-scale feature fusion and selection, further enhancing the model’s ability to represent weather degradation features with diverse scales. Extensive experimental results demonstrate that the proposed method achieves superior performance over existing state-of-the-art approaches on multiple remote sensing image restoration benchmarks, while maintaining a favorable balance between performance improvement and computational efficiency.</p>