【摘要】 针对因数字化处理或者网络传输过程中导致的图片质量不佳的情况，研究旧书重做方法及其应用。以传统的超分辨领域中的插值算法为原型，针对传统的插值算法存在的鲁棒性差、计算量大的问题进行改进，将深度学习方法融入传统的插值算法中，提出了网络模型SRCNN。最后通过在set5+旧书测试集上和其他算法进行对比，得到了SRCNN算法在不同的上采样倍率条件下性能都优于传统插值算法的结果，证明了算法的实用性和优越性。更多还原

【Abstract】 In view of the poor quality of images caused by the process of digital processing or network transmission, the redoing method and its application of old books are studied. With the traditional interpolation algorithm in the super-resolution field, the deep learning method is integrated into the traditional interpolation algorithm, and the network model SRCNN is proposed. Finally,by comparing with other algorithms in set5 and old book test set, the performance of SRCNN algorithm is better than the traditional interpolation algorithm under different upsampling multiplier conditions, which proves the practicality and superiority of the algorithm.更多还原