目的 为了有效地去除多种图像模糊，提高图像质量，提出基于深度强化学习的图像去模糊方法。方法 选用GoPro与DIV2K这2个数据集进行实验，以峰值信噪比（PSNR）和结构相似性（SSIM）为客观评价指标。通过卷积神经网络获得模糊图像的高维特征，利用深度强化学习结合多种CNN去模糊工具建立去模糊框架，将峰值信噪比（PSNR）作为训练奖励评价函数，来选择最优修复策略，逐步对模糊图像进行修复。结果 通过训练与测试，与现有的主流算法相比，文中方法有着更好的主观视觉效果，且PSNR值与SSIM值都有更好的表现。结论 实验结果表明，文中方法能有效地解决图像的高斯模糊和运动模糊等问题，并取得了良好的视觉效果，在图像去模糊领域具有一定的参考价值。

The paper aims to propose an image deblurring method based on deep reinforcement learning to effectively remove multiple image blurs and improve image quality. GoPro and DIV2K datasets were used for experiments. The peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) were used as objective evaluation indicators. The high-dimensional feature of fuzzy image was obtained by convolutional neural network. The deblurring framework was established by deep reinforcement learning combined with a variety of CNN deblurring tools. The peak signal-to-noise ratio (PSNR) was used as the training reward evaluation function to select the optimal restoration strategy and gradually restore the fuzzy image. Through training and testing, compared with the existing mainstream algorithm, the method presented in this paper had a better subjective visual effect; and the PSNR value and SSIM value had better performance. The experimental results show that the method in this paper can effectively solve the problem of Gaussian blur and motion blur of image, and obtain good visual effects. It has certain reference value in image deblurring.