SPH-Net: Hyperspectral Image Super-Resolution via Smoothed Particle Hydrodynamics Modeling

Area of Science:

• Computer Vision
• Image Processing
• Remote Sensing

Background:

• Hyperspectral image (HSI) super-resolution is crucial for applications in remote sensing and aerospace.
• Existing deep learning methods often focus on network architecture, neglecting the dynamic evolution of image pixels during reconstruction.

Purpose of the Study:

• To propose a novel hyperspectral image super-resolution method inspired by smoothed particle hydrodynamics (SPH).
• To develop an SPH-based network (SPH-Net) that addresses HSI super-resolution by considering pixel dynamics.

Main Methods:

• Developed an SPH network (SPH-Net) leveraging smoothed particle hydrodynamics theory for HSI super-resolution.
• Constructed a smooth function based on SPH and designed multiscale smooth convolutions to exploit spectral correlations.
• Applied SPH approximation to discretize the Navier-Stokes equation, guiding pixel motion for improved spatial clarity.

Main Results:

• SPH-Net demonstrated superior performance compared to state-of-the-art methods on three public hyperspectral datasets.
• The method achieved significant improvements in both objective metrics and visual quality of super-resolved HSIs.
• The SPH-guided approach effectively enhanced edge clarity in the spatial domain.

Conclusions:

• The proposed SPH-Net offers a novel and effective approach to hyperspectral image super-resolution by incorporating principles of fluid dynamics.
• The method successfully preserves spectral information while enhancing spatial resolution and edge definition.
• SPH-Net represents a significant advancement in HSI super-resolution techniques for remote sensing and related fields.