The field of image restoration is moving towards more sophisticated and efficient methods, with a focus on leveraging depth information, task-aware prompting, and inter-component correction. Recent developments have shown that incorporating depth guidance can significantly improve restoration quality, while task-aware prompting can enable more accurate and parameter-efficient restoration models. Additionally, addressing inter-component residuals and exposure inconsistencies has been identified as crucial for achieving high-quality image enhancement. Noteworthy papers include:

• A novel Depth-Guided Network for image restoration, which achieves state-of-the-art performance on several benchmarks.
• A parameter-efficient All-in-One image restoration framework that leverages task-aware enhanced prompts to tackle various adverse weather degradations, achieving superior performance with only 2.75M parameters.
• An Inter-correction Retinex model that alleviates inter-component residuals during decomposition and enhancement, outperforming state-of-the-art approaches on low-light benchmark datasets.
• A Wavelet-based Exposure Correction method guided by degradation description, which effectively addresses intra-class variability and achieves significant performance improvements on multiple public datasets.