一种基于小波变换与Mamba的双域融合全色锐化方法及系统
By employing a dual-domain fusion method combining wavelet transform and Mamba, the problems of spectral distortion and insufficient spatial detail in panchromatic sharpening are solved, achieving efficient and stable spectral consistency and detail clarity, which is suitable for global modeling of high-resolution remote sensing images.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU OCEAN UNIV
- Filing Date
- 2025-11-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing panchromatic sharpening methods tend to introduce spectral distortion when enhancing spatial details. Traditional methods are difficult to build robust optimizations for complex scenes, while deep learning methods have high computational overhead and loose cross-modal coupling, making it difficult to achieve efficient global modeling.
A dual-domain fusion method combining wavelet transform and Mamba is adopted. Through decomposition and feature extraction in the frequency and spatial domains, combined with high-frequency detail residual injection and upsampling, cross-modal feature fusion is achieved. Furthermore, Mamba interactive modeling and channel attention reconstruction are used to ensure spectral consistency and detail clarity.
It significantly reduces spectral distortion, preserves spatial details, improves spectral consistency, and achieves efficient global modeling. It is suitable for the computational efficiency and deployment cost of high-resolution remote sensing images, has strong multi-scale representation capabilities, and is stable in training and convergence.
Smart Images
Figure CN121582099B_ABST