A CNN-Transformer Hybrid Feature Descriptor for Optical-SAR Image Registration

This paper presents a CNN-Transformer hybrid feature descriptor for the registration of optical and SAR images. The proposed feature descriptor is composed of a shallow feature extraction module based on CNN and a deep feature extraction module based on Transformer, which are tailored to extract fea...

Full description

Saved in:
Bibliographic Details
Main Authors: Lin, Mingxin, Liu, Binyuan, Liu, Yijun, Wang, Qingsong
Format: Conference Proceeding
Language:English
Subjects:
Adaptive optics
CNN
feature descriptor
Feature extraction
Geoscience and remote sensing
Image registration
Optical imaging
optical-SAR image registration
Radar polarimetry
Transformer
Transformers
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents a CNN-Transformer hybrid feature descriptor for the registration of optical and SAR images. The proposed feature descriptor is composed of a shallow feature extraction module based on CNN and a deep feature extraction module based on Transformer, which are tailored to extract features from different levels. Its effectiveness is demonstrated in experiments with comparison to state-of-the-art methods. Experimental results demonstrated its superior performance compare to the state-of-the-art methods, thus showcasing its efficacy and potential for practical applications.
ISSN:2153-7003
DOI:10.1109/IGARSS52108.2023.10281985