RDRN: Recursively Defined Residual Network for Image Super-Resolution

Deep convolutional neural networks (CNNs) have obtained remarkable performance in single image super-resolution (SISR). However, very deep can suffer from training difficulty and hardly achieve further gain. There are two main trends to solve that problem: improving the network architecture for better propagation of features through large number layers designing an attention mechanism selecting most informative features. Recent SISR solutions propose advanced self-attention mechanisms. constructing a use block efficient way is challenging problem. To address this issue, we general recursively defined residual (RDRB) feature extraction layers. Based on RDRB designed (RDRN), novel which utilizes blocks efficiently. Extensive experiments show proposed model achieves state-of-the-art results several popular benchmarks outperforms previous methods by up 0.43 dB.