基于多模态几何感知与深度引导动态融合的光场超分方法
By employing an end-to-end hierarchical progressive reconstruction framework based on multimodal geometry perception and depth guidance, the problems of consistency and texture detail reconstruction in light field super-resolution results are solved, enabling effective reconstruction of high-resolution light field images and improving the visual quality of light field images and the robustness of the model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENYANG UNIVERSITY OF TECHNOLOGY
- Filing Date
- 2025-12-19
- Publication Date
- 2026-07-17
AI Technical Summary
Existing light field super-resolution techniques have shortcomings in utilizing light field geometric information, fusing spatial-angular features, and accurate upsampling, which leads to problems such as blurred edges, artifacts, and poor angle consistency in the light field super-resolution results.
An end-to-end hierarchical progressive reconstruction framework based on multimodal geometric perception and depth guidance is adopted. Through geometric information estimation, depth-aware EPI feature extraction, depth-guided feature enhancement, multi-scale geometric perception feature fusion, and geometric consistency constraints, a light field super-resolution model is constructed to achieve high-resolution reconstruction of light field images.
It significantly improves the spatial and angular consistency of light field super-resolution results, enhances the accuracy of texture detail reconstruction, improves the visual quality of light field images, and increases the robustness and adaptability of the model.
Smart Images
Figure CN121746188B_ABST
Abstract
Citation Information
Patent Citations
Light field image spatial super-resolution method based on space-angle decoupling mechanism
CN119048351A
Visible light and infrared image fusion method based on cross-modal dynamic collaboration
CN120525735A