Super-Resolution Reconstruction of Remote Sensing Images Using Chaotic Mapping to Optimize Sparse Representation.

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Super-resolution algorithms struggle with noisy remote sensing images, amplifying noise during high-frequency signal recovery.
• Existing methods often use fixed dictionaries, limiting their adaptability to complex image data.

Purpose of the Study:

• To develop a novel super-resolution algorithm for noisy remote sensing images from space cameras, especially for high-speed imaging.
• To address limitations of current algorithms in noise amplification and detail preservation.

Main Methods:

• Employs K-singular value decomposition (K-SVD) for joint training of high- and low-resolution image blocks to create adaptive dictionary pairs.
• Integrates circle chaotic mapping for improved dictionary updating and uses orthogonal matching pursuit (OMP) for sparse coefficient optimization.
• Applies local gradients as constraints to enhance edge details after upscaling and denoising.

Main Results:

• The proposed algorithm effectively filters noise and artifacts from low-resolution remote sensing images.
• Achieves superior visual quality and objective performance metrics, including peak signal-to-noise ratio and information entropy, compared to existing methods.
• Demonstrates high-quality remote sensing image data generation.

Conclusions:

• The novel approach significantly enhances super-resolution for noisy remote sensing images.
• The method offers a robust solution for improving image quality in space-based high-speed imaging systems.
• Validated effectiveness in producing high-fidelity remote sensing imagery.