1: \begin{abstract}

2: Neural-networks based image restoration methods tend to use low-resolution image patches for training.

3: Although higher-resolution image patches can provide more global information, state-of-the-art methods cannot utilize them due to their huge GPU memory usage, as well as the instable training process.

4: However, plenty of studies have shown that global information is crucial for image restoration tasks like image demosaicing and enhancing.

5: In this work, we propose a HighEr-Resolution Network (HERN) to fully learning global information in high-resolution image patches. 

6: To achieve this, the HERN employs two parallel paths to learn image features in two different resolution, respectively.

7: By combining global-aware features and multi-scale features, our HERN is able to learn global information with feasible GPU memory usage.

8: Besides, we introduce a progressive training method to solve the instability issue and accelerate model convergence.

9: On the task of image demosaicing and enhancing, our HERN achieves state-of-the-art performance on the AIM2019 RAW to RGB mapping challenge.

10: The source code of our implementation is available at \url{https://github.com/MKFMIKU/RAW2RGBNet}.

11: % The code will be released soon.

12: \end{abstract}

13: