Generating high quality visible images from SAR images using CNNs

Puyang Wang; Vishal M. Patel

doi:10.1109/RADAR.2018.8378622

Generating high quality visible images from SAR images using CNNs

Puyang Wang
, Vishal M. Patel

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We propose a novel approach for generating high quality visible-like images from Synthetic Aperture Radar (SAR) images using Deep Convolutional Generative Adversarial Network (GAN) architectures. The proposed approach is based on a cascaded network of convolutional neural nets (CNNs) for despeckling and image colorization. The cascaded structure results in faster convergence during training and produces high quality visible images from the corresponding SAR images. Experimental results on both simulated and real SAR images show that the proposed method can produce visible-like images better compared to the recent state-of-the-art deep learning-based methods.

Original language	English (US)
Title of host publication	2018 IEEE Radar Conference, RadarConf 2018
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	570-575
Number of pages	6
ISBN (Electronic)	9781538641675
DOIs	https://doi.org/10.1109/RADAR.2018.8378622
State	Published - Jun 8 2018
Externally published	Yes
Event	2018 IEEE Radar Conference, RadarConf 2018 - Oklahoma City, United States Duration: Apr 23 2018 → Apr 27 2018

Publication series

Name	2018 IEEE Radar Conference, RadarConf 2018

Conference

Conference	2018 IEEE Radar Conference, RadarConf 2018
Country/Territory	United States
City	Oklahoma City
Period	4/23/18 → 4/27/18

Keywords

colorization
despeckling
Synthetic aperture radar image

ASJC Scopus subject areas

Computer Networks and Communications
Signal Processing
Instrumentation

Access to Document

10.1109/RADAR.2018.8378622

Cite this

Wang, P & Patel, VM 2018, Generating high quality visible images from SAR images using CNNs. in 2018 IEEE Radar Conference, RadarConf 2018. 2018 IEEE Radar Conference, RadarConf 2018, Institute of Electrical and Electronics Engineers Inc., pp. 570-575, 2018 IEEE Radar Conference, RadarConf 2018, Oklahoma City, United States, 4/23/18. https://doi.org/10.1109/RADAR.2018.8378622

@inproceedings{b8098edab637451ca3fd80e6f7c771c7,

title = "Generating high quality visible images from SAR images using CNNs",

abstract = "We propose a novel approach for generating high quality visible-like images from Synthetic Aperture Radar (SAR) images using Deep Convolutional Generative Adversarial Network (GAN) architectures. The proposed approach is based on a cascaded network of convolutional neural nets (CNNs) for despeckling and image colorization. The cascaded structure results in faster convergence during training and produces high quality visible images from the corresponding SAR images. Experimental results on both simulated and real SAR images show that the proposed method can produce visible-like images better compared to the recent state-of-the-art deep learning-based methods.",

keywords = "colorization, despeckling, Synthetic aperture radar image",

author = "Puyang Wang and Patel, \{Vishal M.\}",

note = "Publisher Copyright: {\textcopyright} 2018 IEEE.; 2018 IEEE Radar Conference, RadarConf 2018 ; Conference date: 23-04-2018 Through 27-04-2018",

year = "2018",

month = jun,

day = "8",

doi = "10.1109/RADAR.2018.8378622",

language = "English (US)",

series = "2018 IEEE Radar Conference, RadarConf 2018",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "570--575",

booktitle = "2018 IEEE Radar Conference, RadarConf 2018",

}

TY - GEN

T1 - Generating high quality visible images from SAR images using CNNs

AU - Wang, Puyang

AU - Patel, Vishal M.

PY - 2018/6/8

Y1 - 2018/6/8

N2 - We propose a novel approach for generating high quality visible-like images from Synthetic Aperture Radar (SAR) images using Deep Convolutional Generative Adversarial Network (GAN) architectures. The proposed approach is based on a cascaded network of convolutional neural nets (CNNs) for despeckling and image colorization. The cascaded structure results in faster convergence during training and produces high quality visible images from the corresponding SAR images. Experimental results on both simulated and real SAR images show that the proposed method can produce visible-like images better compared to the recent state-of-the-art deep learning-based methods.

AB - We propose a novel approach for generating high quality visible-like images from Synthetic Aperture Radar (SAR) images using Deep Convolutional Generative Adversarial Network (GAN) architectures. The proposed approach is based on a cascaded network of convolutional neural nets (CNNs) for despeckling and image colorization. The cascaded structure results in faster convergence during training and produces high quality visible images from the corresponding SAR images. Experimental results on both simulated and real SAR images show that the proposed method can produce visible-like images better compared to the recent state-of-the-art deep learning-based methods.

KW - colorization

KW - despeckling

KW - Synthetic aperture radar image

UR - https://www.scopus.com/pages/publications/85049966964

UR - https://www.scopus.com/pages/publications/85049966964#tab=citedBy

U2 - 10.1109/RADAR.2018.8378622

DO - 10.1109/RADAR.2018.8378622

M3 - Conference contribution

AN - SCOPUS:85049966964

T3 - 2018 IEEE Radar Conference, RadarConf 2018

SP - 570

EP - 575

BT - 2018 IEEE Radar Conference, RadarConf 2018

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2018 IEEE Radar Conference, RadarConf 2018

Y2 - 23 April 2018 through 27 April 2018

ER -

Generating high quality visible images from SAR images using CNNs

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this