A dual-stream remote sensing image fusion method based on progressive frequency perception
By constructing a progressive frequency-aware dual-stream remote sensing image fusion method, and employing dynamic low-pass filtering and frequency domain cross-attention mechanism, the problems of insufficient separation of high and low frequency features and cross-modal information interaction in existing technologies are solved, thereby improving the remote sensing image fusion effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2026-04-20
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning remote sensing image fusion models struggle to achieve accurate separation and refined processing of high and low frequency features, and lack frequency domain-level cross-modal information interaction. This makes it difficult to simultaneously enhance spatial details and maintain spectral fidelity in fused images, hindering further improvements in fusion performance from existing technologies.
A dual-stream remote sensing image fusion method based on progressive frequency awareness is constructed. It adopts a cascaded network architecture without encoding and decoding, realizes high- and low-frequency feature separation and enhancement through dynamic low-pass filtering, and designs a frequency domain cross-attention mechanism for cross-modal information interaction to maintain high-resolution feature processing.
It significantly improved the remote sensing image fusion effect, enhanced the spatial details and spectral consistency of the fused images, and achieved an improvement in image quality.
Smart Images
Figure CN122415347A_ABST