[Home ] [Archive]   [ فارسی ]  
:: Main :: About :: Current Issue :: Archive :: Search :: Submit :: Contact ::
Main Menu
Home::
Journal Information::
Articles archive::
For Authors::
For Reviewers::
Registration::
Contact us::
Site Facilities::
::
Search in website

Advanced Search
..
Receive site information
Enter your Email in the following box to receive the site news and information.
..
:: Volume 11, Issue 2 (9-2023) ::
jgit 2023, 11(2): 41-62 Back to browse issues page
Spatial resolution improvement of the multispectral satellite images in the absence of high spatial resolution images through integration of image fusion and deep learning methods
Hamid Amini Amirkolaee , Alireza Safdarinezhad * , Hamed Amini Amirkolaee
Tafresh University
Abstract:   (1136 Views)
Improving the spatial resolution of multispectral images is one of the common pre-processing steps in reaching the maximum potential of these images in remote sensing applications. The presence of images with higher spatial resolution along with multispectral images allows the process of improving spatial resolution to be performed through image pan-sharpening methods. The lack of simultaneous panchromatic image sensors in the satellite platforms imposes challenges related to co-registration and asynchronies when using images of different satellite sensors in the process of image pan-sharpening. In such a situation, super-resolution techniques are considered as alternative approaches to improve spatial resolution. Using the generative adversarial network (GAN) is one of the effective methods in this field that require the existence of multiple training data. Generally, it is not possible to prepare two satellite images with the same spectral resolution and different spatial resolution from a specific region that is required for training the network. Therefore, in this research, an approach with two main steps is designed to improve the spatial resolution of multispectral images. In the first step, a deep super-resolution generative adversarial network is used to improve the resolution of the true color composition of multispectral images. A boosting strategy is exploited to deeply train the GAN network using the resampled images extracted from the Google-Earth. In the second step, the spectral contents are added to the super-resolution images through the traditional pan-sharpening method. The results demonstrated that the proposed approach improved the spatial resolution of multispectral images by 32.85% better than the best comparative method and maintained the spectral content without the need to provide extensive training data.
 
Keywords: Image fusion, Super-resolution, Multispectral images, Deep learning, Boost learning
Full-Text [PDF 2025 kb]   (183 Downloads)    
Type of Study: Research | Subject: RS
Received: 2023/06/10 | Accepted: 2023/08/19 | ePublished ahead of print: 2023/08/28 | Published: 2023/10/10
Send email to the article author


XML   Persian Abstract   Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

Amini Amirkolaee H, Safdarinezhad A, Amini Amirkolaee H. Spatial resolution improvement of the multispectral satellite images in the absence of high spatial resolution images through integration of image fusion and deep learning methods. jgit 2023; 11 (2) :41-62
URL: http://jgit.kntu.ac.ir/article-1-919-en.html


Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Volume 11, Issue 2 (9-2023) Back to browse issues page
نشریه علمی-پژوهشی مهندسی فناوری اطلاعات مکانی Engineering Journal of Geospatial Information Technology
Persian site map - English site map - Created in 0.06 seconds with 35 queries by YEKTAWEB 4645