Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Fusion-Guided Inception Network for Hyperspectral Image Super-Resolution

Usman Muhammad, Jorma Laaksonen

TL;DR

The paper addresses the challenge of hyperspectral single-image super-resolution without relying on aligned high-resolution auxiliary images. It introduces the Fusion-Guided Inception Network (FGIN), which integrates spectral–spatial fusion, Inception-like multi-scale feature extraction, multi-scale fusion, and an optimized upsampling module to reconstruct high-resolution HSIs from a single LR input. Key contributions include band grouping to manage spectral dimensionality, a four-module FGIn architecture, and detailed ablation showing the method’s efficiency (≈1.07M parameters) and effectiveness on public datasets. Experiments on PaviaC and PaviaU across scale factors of 2, 4, and 8 demonstrate competitive performance, highlighting the method’s practicality for real-world hyperspectral SR tasks with limited auxiliary information.

Abstract

The fusion of low-spatial-resolution hyperspectral images (HSIs) with high-spatial-resolution conventional images (e.g., panchromatic or RGB) has played a significant role in recent advancements in HSI super-resolution. However, this fusion process relies on the availability of precise alignment between image pairs, which is often challenging in real-world scenarios. To mitigate this limitation, we propose a single-image super-resolution model called the Fusion-Guided Inception Network (FGIN). Specifically, we first employ a spectral-spatial fusion module to effectively integrate spectral and spatial information at an early stage. Next, an Inception-like hierarchical feature extraction strategy is used to capture multiscale spatial dependencies, followed by a dedicated multi-scale fusion block. To further enhance reconstruction quality, we incorporate an optimized upsampling module that combines bilinear interpolation with depthwise separable convolutions. Experimental evaluations on two publicly available hyperspectral datasets demonstrate the competitive performance of our method.

A Fusion-Guided Inception Network for Hyperspectral Image Super-Resolution

TL;DR

The paper addresses the challenge of hyperspectral single-image super-resolution without relying on aligned high-resolution auxiliary images. It introduces the Fusion-Guided Inception Network (FGIN), which integrates spectral–spatial fusion, Inception-like multi-scale feature extraction, multi-scale fusion, and an optimized upsampling module to reconstruct high-resolution HSIs from a single LR input. Key contributions include band grouping to manage spectral dimensionality, a four-module FGIn architecture, and detailed ablation showing the method’s efficiency (≈1.07M parameters) and effectiveness on public datasets. Experiments on PaviaC and PaviaU across scale factors of 2, 4, and 8 demonstrate competitive performance, highlighting the method’s practicality for real-world hyperspectral SR tasks with limited auxiliary information.

Abstract

The fusion of low-spatial-resolution hyperspectral images (HSIs) with high-spatial-resolution conventional images (e.g., panchromatic or RGB) has played a significant role in recent advancements in HSI super-resolution. However, this fusion process relies on the availability of precise alignment between image pairs, which is often challenging in real-world scenarios. To mitigate this limitation, we propose a single-image super-resolution model called the Fusion-Guided Inception Network (FGIN). Specifically, we first employ a spectral-spatial fusion module to effectively integrate spectral and spatial information at an early stage. Next, an Inception-like hierarchical feature extraction strategy is used to capture multiscale spatial dependencies, followed by a dedicated multi-scale fusion block. To further enhance reconstruction quality, we incorporate an optimized upsampling module that combines bilinear interpolation with depthwise separable convolutions. Experimental evaluations on two publicly available hyperspectral datasets demonstrate the competitive performance of our method.
Paper Structure (12 sections, 11 equations, 1 figure, 2 tables)

This paper contains 12 sections, 11 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Band Grouping
  4. Spectral-Spatial Fusion
  5. Inception-like Blocks
  6. Multi-Scale Fusion Block
  7. Upsampling Block
  8. Experimental setup
  9. Implementation
  10. Ablation Study
  11. Comparison with State-of-the-Art Methods
  12. Conclusion

Figures (1)

  • Figure 1: Overview of the proposed FGIN model. The architecture consists of: (1) a Spectral-Spatial Fusion module (left); (2) Inception-like multiscale feature extractors (center) for hierarchical representation learning; (3) a Multi-Scale Fusion block for feature refinement; and (4) an Upsampling block (right) that reconstructs high-resolution (HR) hyperspectral images.