【摘要】 目前，失真图像在图像获取、转存和传输过程中发生图像信息丢失问题日益严重，然而目前很多基于深度学习的图像修复方法都需要大量的数据集支持才能实现图像修复的高复原度这一指标，严重耗费时间和资源。提出一种新型循环卷积融合神经网络模型来实现图像恢复，将原本的卷积层重构后获得的特征图导入循环模型进行信息获取，并通过U-net网络结构实现重建以得到结果。在失真数据集TID2008和TID2013上的实验结果证明，数据集较少的情况下，修复效果相对于传统方法更佳。更多还原

【Abstract】 Recently,the problem of image information loss in the process of image acquisition,storage and transmission of distorted images is becoming increasingly serious,but many image repair methods based on deep learning require a large number of data sets to achieve high resilience of image repair,which consumes time and resources. In this paper,a new recurrent convolutional fusion neural network model is proposed to realize image recovery,and the feature map obtained after the original convolutional layer reconstruction is imported into the recurrent model for information acquisition,and the results are obtained through the reconstruction of the U-net network structure. Experimental results on the distortion datasets TID2008 and TID2013 show that the repair effect is better than that of traditional methods when the data set is small.更多还原