TY - GEN
T1 - Orthogonally regularized deep networks for image super-resolution
AU - Guo, Tiantong
AU - Mousavi, Hojjat S.
AU - Monga, Vishal
N1 - Funding Information:
This work is supported by an NSF CAREER Award to V. Monga.
Publisher Copyright:
© 2018 IEEE.
Copyright:
Copyright 2018 Elsevier B.V., All rights reserved.
PY - 2018/9/10
Y1 - 2018/9/10
N2 - Deep learning methods, in particular trained Convolutional Neural Networks (CNNs) have recently been shown to produce compelling state-of-the-art results for single image Super-Resolution (SR). Invariably, a CNN is learned to map the low resolution (LR) image to its corresponding high resolution (HR) version in the spatial domain. Aiming for faster inference and more efficient solutions than solving the SR problem in the spatial domain, we propose a novel network structure for learning the SR mapping function in an image transform domain, specifically the Discrete Cosine Transform (DCT). As a first contribution, we show that DCT can be integrated into the network structure as a Convolutional DCT (CDCT) layer. We further extend the network to allow the CDCT layer to become trainable (i.e. optimizable). Because this layer represents an image transform, we enforce pairwise orthogonality constraints on the individual basis functions/filters. This Orthogonally Regularized Deep SR network (ORDSR) simplifies the SR task by taking advantage of image transform domain while adapting the design of transform basis to the training image set. Experimental results show ORDSR achieves state-of-the-art SR image quality with fewer parameters than most of the deep CNN methods.
AB - Deep learning methods, in particular trained Convolutional Neural Networks (CNNs) have recently been shown to produce compelling state-of-the-art results for single image Super-Resolution (SR). Invariably, a CNN is learned to map the low resolution (LR) image to its corresponding high resolution (HR) version in the spatial domain. Aiming for faster inference and more efficient solutions than solving the SR problem in the spatial domain, we propose a novel network structure for learning the SR mapping function in an image transform domain, specifically the Discrete Cosine Transform (DCT). As a first contribution, we show that DCT can be integrated into the network structure as a Convolutional DCT (CDCT) layer. We further extend the network to allow the CDCT layer to become trainable (i.e. optimizable). Because this layer represents an image transform, we enforce pairwise orthogonality constraints on the individual basis functions/filters. This Orthogonally Regularized Deep SR network (ORDSR) simplifies the SR task by taking advantage of image transform domain while adapting the design of transform basis to the training image set. Experimental results show ORDSR achieves state-of-the-art SR image quality with fewer parameters than most of the deep CNN methods.
UR - http://www.scopus.com/inward/record.url?scp=85054289002&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85054289002&partnerID=8YFLogxK
U2 - 10.1109/ICASSP.2018.8462555
DO - 10.1109/ICASSP.2018.8462555
M3 - Conference contribution
AN - SCOPUS:85054289002
SN - 9781538646588
T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
SP - 1463
EP - 1467
BT - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018
Y2 - 15 April 2018 through 20 April 2018
ER -