Multi-attention feature fusion network for lightweight image super-resolution

Abstract

Aiming at the problems of image super-resolution reconstruction method based on convolutional neural network, such as complex structure, huge parameter amount, and slow reconstruction speed, this paper proposes a multi-attention feature fusion network for lightweight image super-resolution. The channel attention mechanism and pixel attention mechanism are used to fully extract image feature information and improve the feature extraction ability of the network. At the same time, the use of depthwise convolution effectively reduces the amount of network parameters and calculations. The experimental results show that under the premise that the reconstruction performance of the network is competitive, the network is lighter and the reconstruction speed is further improved.