elib
DLR-Header
DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Aerial and Satellite Image Enhancement with Super Resolution using Deep Learning

Jangir, Sandeep Kumar (2020) Aerial and Satellite Image Enhancement with Super Resolution using Deep Learning. Master's, Technical University of Munich.

This is the latest version of this item.

Full text not available from this repository.

Abstract

Single-Image Super Resolution (SISR) is a method to convert a low-resolution (LR) image to a high-resolution (HR) image. As the applications for aerial and satellite images arebecoming popular, there is a need for super resolution algorithms to enhance the quality the images to perform various image processing tasks such as object detection, tracking, image segmentation, and etc. Even though the basic interpolation techniques could be used to increase the spatial dimensions of an image, they are not able to reconstruct the amount of details. The aim of this thesis is to improve the SR results for aerial and satellite images using deep learning-based algorithms such as CNNs and GANs. As the resulting SR image will be further used to perform tiny lane and dense object segmentation from aerial and satellite images, thus the developed algorithm must not only enhance the quality of the images but also retrieve fine information. Training a SR algorithm requires a LR image as input and HR image as the ground-truth.The LR image is created by downsampling the HR image and since there are several ways to downsample a HR image, a comparison is made between these methods and the best method is used to create the LR images for the experiments. After creating the LR-HR image pairs for training, the proposed method is trained on aerial and satellite images respectively and their results are tabulated. Cross testing is performed between them in order to check their generic behaviours on aerial and satellite images. Even though increasing the resolution of a satellite images produce a better quality image, the perceptual quality of the resulting image still lacks the details of an aerial image. An experiment is perform to see if a satellite image can be converted into an aerial image while simultaneously increasing its spatial resolution.

Item URL in elib:https://elib.dlr.de/138151/
Document Type:Thesis (Master's)
Additional Information:This master thesis was supervised by Seyed Majid Azimi and Dr. Reza Bahmanyar.
Title:Aerial and Satellite Image Enhancement with Super Resolution using Deep Learning
Authors:
AuthorsInstitution or Email of AuthorsAuthor's ORCID iD
Jangir, Sandeep Kumarsandeep.jangir (at) dlr.deUNSPECIFIED
Date:2020
Refereed publication:No
Open Access:No
Gold Open Access:No
In SCOPUS:No
In ISI Web of Science:No
Number of Pages:111
Status:Published
Keywords:Single-image Super Resolution (SISR), image enhancement, aerial and satelliteimages, convolutional neural networks, ESRGAN, SAN, interpolation
Institution:Technical University of Munich
Department:Faculty of Computer Science
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Transport
HGF - Program Themes:Road Transport
DLR - Research area:Transport
DLR - Program:V ST Straßenverkehr
DLR - Research theme (Project):V - D.MoVe
Location: Oberpfaffenhofen
Institutes and Institutions:Remote Sensing Technology Institute > Photogrammetry and Image Analysis
Deposited By: Azimi, Seyedmajid
Deposited On:27 Nov 2020 10:11
Last Modified:01 Dec 2020 11:14

Available Versions of this Item

  • Aerial and Satellite Image Enhancement with Super Resolution using Deep Learning. (deposited 27 Nov 2020 10:11) [Currently Displayed]

Repository Staff Only: item control page

Browse
Search
Help & Contact
Information
electronic library is running on EPrints 3.3.12
Copyright © 2008-2017 German Aerospace Center (DLR). All rights reserved.