PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss
In the literature of pan-sharpening based on neural networks, high resolution multispectral images as ground-truth labels generally are unavailable. To tackle the issue, a common method is to degrade original images into a lower resolution space for supervised training under the Wald’s protocol. In...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-07-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/12/14/2318 |
id |
doaj-bca0ba2d7f66493eb38fa425a3939719 |
---|---|
record_format |
Article |
spelling |
doaj-bca0ba2d7f66493eb38fa425a39397192020-11-25T03:01:47ZengMDPI AGRemote Sensing2072-42922020-07-01122318231810.3390/rs12142318PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual LossChangsheng Zhou0Jiangshe Zhang1Junmin Liu2Chunxia Zhang3Rongrong Fei4Shuang Xu5The School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaThe School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaThe School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaThe School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaThe School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaThe School of Mathematics and Statistics, Xi’an Jiaotong University, No.28 Xianning West Road, Xi’an 710049, ChinaIn the literature of pan-sharpening based on neural networks, high resolution multispectral images as ground-truth labels generally are unavailable. To tackle the issue, a common method is to degrade original images into a lower resolution space for supervised training under the Wald’s protocol. In this paper, we propose an unsupervised pan-sharpening framework, referred to as “perceptual pan-sharpening”. This novel method is based on auto-encoder and perceptual loss, and it does not need the degradation step for training. For performance boosting, we also suggest a novel training paradigm, called “first supervised pre-training and then unsupervised fine-tuning”, to train the unsupervised framework. Experiments on the QuickBird dataset show that the framework with different generator architectures could get comparable results with the traditional supervised counterpart, and the novel training paradigm performs better than random initialization. When generalizing to the IKONOS dataset, the unsupervised framework could still get competitive results over the supervised ones.https://www.mdpi.com/2072-4292/12/14/2318pan-sharpeningperceptual lossauto-encodergenerative adversarial networksunsupervised learning |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Changsheng Zhou Jiangshe Zhang Junmin Liu Chunxia Zhang Rongrong Fei Shuang Xu |
spellingShingle |
Changsheng Zhou Jiangshe Zhang Junmin Liu Chunxia Zhang Rongrong Fei Shuang Xu PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss Remote Sensing pan-sharpening perceptual loss auto-encoder generative adversarial networks unsupervised learning |
author_facet |
Changsheng Zhou Jiangshe Zhang Junmin Liu Chunxia Zhang Rongrong Fei Shuang Xu |
author_sort |
Changsheng Zhou |
title |
PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss |
title_short |
PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss |
title_full |
PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss |
title_fullStr |
PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss |
title_full_unstemmed |
PercepPan: Towards Unsupervised Pan-Sharpening Based on Perceptual Loss |
title_sort |
perceppan: towards unsupervised pan-sharpening based on perceptual loss |
publisher |
MDPI AG |
series |
Remote Sensing |
issn |
2072-4292 |
publishDate |
2020-07-01 |
description |
In the literature of pan-sharpening based on neural networks, high resolution multispectral images as ground-truth labels generally are unavailable. To tackle the issue, a common method is to degrade original images into a lower resolution space for supervised training under the Wald’s protocol. In this paper, we propose an unsupervised pan-sharpening framework, referred to as “perceptual pan-sharpening”. This novel method is based on auto-encoder and perceptual loss, and it does not need the degradation step for training. For performance boosting, we also suggest a novel training paradigm, called “first supervised pre-training and then unsupervised fine-tuning”, to train the unsupervised framework. Experiments on the QuickBird dataset show that the framework with different generator architectures could get comparable results with the traditional supervised counterpart, and the novel training paradigm performs better than random initialization. When generalizing to the IKONOS dataset, the unsupervised framework could still get competitive results over the supervised ones. |
topic |
pan-sharpening perceptual loss auto-encoder generative adversarial networks unsupervised learning |
url |
https://www.mdpi.com/2072-4292/12/14/2318 |
work_keys_str_mv |
AT changshengzhou perceppantowardsunsupervisedpansharpeningbasedonperceptualloss AT jiangshezhang perceppantowardsunsupervisedpansharpeningbasedonperceptualloss AT junminliu perceppantowardsunsupervisedpansharpeningbasedonperceptualloss AT chunxiazhang perceppantowardsunsupervisedpansharpeningbasedonperceptualloss AT rongrongfei perceppantowardsunsupervisedpansharpeningbasedonperceptualloss AT shuangxu perceppantowardsunsupervisedpansharpeningbasedonperceptualloss |
_version_ |
1724691993087442944 |