Abstract:Recent research on single-image super-resolution (SISR) using deep convolutional neural networks has made a breakthrough and achieved tremendous performance. Despite their significant progress, numerous convolutional neural networks (CNN) are limited in practical applications, owing to the requirement of the heavy computational cost of the model. This paper proposes a multi-path network for SISR, known as multi-path deep CNN with residual inception network for single image super-resolution. In detail, a residual/ResNet block with an Inception block supports the main framework of the entire network architecture. In addition, remove the batch normalization layer from the residual network (ResNet) block and max-pooling layer from the Inception block to further reduce the number of parameters to preventing the over-fitting problem during the training. Moreover, a conventional rectified linear unit (ReLU) is replaced with Leaky ReLU activation function to speed up the training process. Specifically, we propose a novel two upscale module, which adopts three paths to upscale the features by jointly using deconvolution and upsampling layers, instead of using single deconvolution layer or upsampling layer alone. The extensive experimental results on image super-resolution (SR) using five publicly available test datasets, which show that the proposed model not only attains the higher score of peak signal-to-noise ratio/structural similarity index matrix (PSNR/SSIM) but also enables faster and more efficient calculations against the existing image SR methods. For instance, we improved our method in terms of overall PSNR on the SET5 dataset with challenging upscale factor 8× as 1.88 dB over the baseline bicubic method and reduced computational cost in terms of number of parameters 62% by deeply-recursive convolutional neural network (DRCN) method.

Speedy and accurate image super-resolution via deeply recursive CNN with skip connection and network in network.

Lightweight Image Super-Resolution Network Using 3D Convolutional Neural Networks

Fast and Accurate Single Image Super-Resolution Via an Energy-Aware Improved Deep Residual Network.

Epistemic-Uncertainty-Based Divide-and-Conquer Network for Single-Image Super-Resolution

Fast and Accurate Image Super Resolution by Deep CNN with Skip Connection and Network in Network

NoUCSR: Efficient Super-Resolution Network Without Upsampling Convolution.

Symmetrical Residual Connections for Single Image Super-Resolution

A Single Super-Resolution Method Via Deep Cascade Network

Spatial and Channel Aggregation Network for Lightweight Image Super-Resolution

Sub-Pixel Convolutional Neural Network for Image Super-Resolution Reconstruction

Deeply Recursive Low- and High-Frequency Fusing Networks for Single Image Super-Resolution

Memory Recursive Network for Single Image Super-Resolution

Iterative Network for Image Super-Resolution

Deep Neural Network for Fast and Accurate Single Image Super-Resolution via Channel-Attention-based Fusion of Orientation-aware Features

HNSR: Highway Networks Based Deep Convolutional Neural Networks Model for Single Image Super-Resolution

Fast and Accurate Single Image Super-Resolution via Information Distillation Network

Lightweight and Accurate Single Image Super-Resolution with Channel Segregation Network.

Lightweight image super-resolution with enhanced CNN

Multi-Path Deep CNN with Residual Inception Network for Single Image Super-Resolution

A New Convolutional Neural Network for Super-Resolution by Global and Local Residual

Single-Image Super-Resolution Reconstruction based on the Differences of Neighboring Pixels