Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

MCFRNet: Multi-path contextual feature refinement network for hyperspectral image classification.

Xiaoqing Wan1, Ziqi Sun2, Yupeng He2

  • 1Hengyang Normal University, College of Computer Science and Technology, Hengyang, 421002, China; Hunan Provincial Key Laboratory of Intelligent Information Processing and Application, Hengyang, 421002, China.

Neural Networks : the Official Journal of the International Neural Network Society
|July 9, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Midline signaling regulates kidney positioning but not nephrogenesis through Shh.

Developmental biology·2010
Same author

Y chromosomal STR polymorphism in northern Chinese populations.

Biological research·2010
Same author

Construction of NF-κB-targeting RNAi adenovirus vector and the effect of NF-κB pathway on proliferation and apoptosis of vascular endothelial cells.

Molecular biology reports·2010
Same author

Hearing evaluation of intratympanic methylprednisolone perfusion for refractory sudden sensorineural hearing loss.

Otolaryngology--head and neck surgery : official journal of American Academy of Otolaryngology-Head and Neck Surgery·2010
Same author

Hydrothermal synthesis of nanostructures Bi12TiO20 and their photocatalytic activity on acid orange 7 under visible light.

Chemosphere·2010
Same author

An anticancer drug delivery system based on surfactant-templated mesoporous silica nanoparticles.

Biomaterials·2010

This study introduces MCFRNet, a novel network for hyperspectral image classification. It efficiently refines spectral-spatial features, achieving high accuracy with low computational cost.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Remote Sensing

Background:

  • Hyperspectral image (HSI) classification faces challenges in balancing contextual information with local details and computational efficiency.
  • Convolutional Neural Networks (CNNs) excel at local features but have limited receptive fields.
  • Transformers capture long-range dependencies but are computationally expensive and may lose local details.

Purpose of the Study:

  • To propose a lightweight multi-path contextual feature refinement network (MCFRNet) for HSI classification.
  • To address the limitations of existing CNN and Transformer-based methods in HSI analysis.
  • To achieve high classification accuracy with reduced computational complexity.

Main Methods:

  • Designed the HEFE module using depthwise and depthwise separable convolutions to expand the receptive field cost-effectively.
Keywords:
Convolutional neural network (CNN)Feature fusionHyperspectral image (HSI)Self-aware coordinated attention

Related Experiment Videos

  • Integrated the MCFR module with multi-layer feature reconstruction (MFR) and multi-group contextual feature aggregation (MCFA) for unified feature refinement.
  • Employed the SACA module, combining channel and global spatial attention, to enhance spectral channels and spatial correlations.

    • Main Results:

    • MCFRNet demonstrated competitive classification accuracy on four public HSI datasets.
    • The proposed network achieved high accuracy with a low parameter count.
    • MCFRNet exhibited low computational complexity compared to existing methods.

    Conclusions:

    • MCFRNet effectively refines spectral-spatial features for HSI classification.
    • The lightweight architecture offers a promising solution for efficient and accurate HSI analysis.
    • The proposed method balances contextual modeling and local detail preservation effectively.