Loading…
OptiLCD: an optimal lossless compression and denoising technique for satellite images using hybrid optimization and deep learning techniques
Geoinformation from satellite images is used for a variety of earth science applications. Because of the limitations of optics and sensor technology and the high cost of Earth observation satellites, spatial and spectral resolution are not always at the desired high level. In instruction to recover...
Saved in:
Published in: | Soft computing (Berlin, Germany) Germany), 2023-12, Vol.27 (24), p.18605-18622 |
---|---|
Main Authors: | , |
Format: | Article |
Language: | English |
Subjects: | |
Citations: | Items that this one cites |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Geoinformation from satellite images is used for a variety of earth science applications. Because of the limitations of optics and sensor technology and the high cost of Earth observation satellites, spatial and spectral resolution are not always at the desired high level. In instruction to recover the spatial and spectral values of satellite images, optimization algorithms need to be developed. For satellite image super-resolution, several deep learning techniques have been proposed recently. The drawback is that these models may not perform adequately in a wide range of real-world situations because they are trained using synthetic data generated through a limited amount of real data collected from satellites. Due to the one-dimensional structure of the convolutional kernels in the existing learning technique, diversity is lacking and the propagation of information is affected. As satellite images are being archived continually along with their increasing use in various applications, there is an increasing need for effective compression methods. In this study, we address the joint optimization problems, i.e., satellite image denoising and compression. We propose optimal lossless compression and denoising technique (OptiLCD) using hybrid optimization and deep learning techniques for the satellite images. To remove unwanted artifacts from satellite images, hybrid teaching learning-induced Harris-hawks optimization (hybridTL-HHO) algorithm was used. For compression of satellite images, we use convolutional neural networks (CNNs). A new method, improved beetle swarm optimization (IBSO), has been introduced to enhance the performance of CNNs that compute optimal weight parameters to construct encoder–decoder lattices. To conclude, using open-source benchmark datasets, we validate the performance of our OptiLCD model. As far as PSNR, SSIM, and compression ratio are concerned, OptiLCD model can be compared to the standing state-of-the-art replicas. |
---|---|
ISSN: | 1432-7643 1433-7479 |
DOI: | 10.1007/s00500-023-09361-9 |