Open AccessImage High Frequency Information Restoration Algorithm Based on Deep Learning
Author(s) -
Liang Ma,
Zhou Rong-ji,
Ke Zhang
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/769/3/032008
Subject(s) - artificial intelligence , computer science , residual , convolutional neural network , feature (linguistics) , image (mathematics) , compensation (psychology) , deep learning , pattern recognition (psychology) , artificial neural network , algorithm , computer vision , iterative reconstruction , image restoration , image processing , psychology , philosophy , linguistics , psychoanalysis
In order to effectively recover the high-frequency information of images, this paper proposes a deep neural network based on feature compensation to reconstruct super-resolution images. In this method, intensive deep convolutional neural network and residual network are combined, and the high-frequency information of the original image is extracted separately after up sampling and fused with the reconstructed super-resolution image to form high-frequency feature compensation, which can improve the image quality. Through experimental comparison, the effect of the super resolution image reconstructed by the proposed algorithm is improved by about 1dB compared with that reconstructed by the SRCNN algorithm.