Multi-source remote sensing image spatio-temporal fusion method based on edge alignment
By creating probabilistic maps and comparing multi-resolution similarity at boundary locations, and dynamically updating edge alignment information, the problems of false edges and cloud filling errors caused by changes in ground feature edges are solved, enabling adaptive fusion and accurate restoration of remote sensing images.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUZHOU UNIV
- Filing Date
- 2026-05-28
- Publication Date
- 2026-07-10
AI Technical Summary
Existing spatiotemporal fusion methods based on edge alignment for multi-source remote sensing images fail when dealing with dynamic changes in the edges of ground features over time. This leads to the failure of static edge alignment, the generation of false edges, and incorrect cloud filling directions, resulting in distortion of the fused images and inaccurate quantitative inversion.
By establishing a boundary preservation probability map for each boundary location and combining structural similarity comparisons at multiple resolution levels, the boundaries that need to be updated are dynamically identified. Alternative boundary information is extracted from recent cloudless high-resolution images and smoothly fused. In cloud-occluded areas, the filling direction weights are adjusted in reverse according to boundary differences, thereby achieving dynamic adaptive updating of edge alignment information.
It effectively solves the alignment deviation problem caused by dynamic changes in the edges of ground features, ensures that the edge information truly reflects the current orientation of ground features, avoids false edge generation and cloud area filling errors, and improves the fidelity of fused images and the reliability of quantitative applications.
Smart Images

Figure CN122367818A_ABST