Deep learning-based Synthetic Aperture Radar (SAR) image despeckling
This report proposes a method of reducing the inherent speckled nature of synthetic aperture radar (SAR) images by integrating optically guided despeckled images with images despeckled using deep learning-based methods applied through a segmentation map utilizing deep learning-based model designated...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/167038 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This report proposes a method of reducing the inherent speckled nature of synthetic aperture radar (SAR) images by integrating optically guided despeckled images with images despeckled using deep learning-based methods applied through a segmentation map utilizing deep learning-based model designated U-Net.
The method of speckle reductions is widely used in SAR image process to improve the quality of SAR images. Many despeckling methods have been discussed to reduce the speckle noise and each method has its own advantages and disadvantages, however there will always be an exchange between speckle reduction and the retention of details in the resultant despeckled SAR images.
Therefore, this report aims to explore the possibility of mitigating the fore-mentioned penalty through the coalescence of two methods of despeckling with complimenting strengths, namely the “guided patch-wise nonlocal SAR despeckling” and “SAR image despeckling using a convolutional neural network”, with the use of sematic segmentation applied by a deep learning U-Net model. |
|---|