面向单幅图像超分辨率的稳定特征增强方法、装置和设备
By using a parallel hybrid enhancement module in a deep backbone network, combined with adaptive subspace partitioning and global prior injection from a learnable prototype library, the modeling challenges of local texture restoration and global long-range dependencies in single-image super-resolution are solved, achieving efficient and stable feature enhancement results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN POLICE ACAD
- Filing Date
- 2026-04-02
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to balance local texture restoration and global long-range dependency modeling in single-image super-resolution reconstruction, and suffer from unstable feature representation and high computational complexity.
Feature enhancement is achieved by using a parallel hybrid enhancement module in a deep backbone network. Through local stable propagation of adaptive subspace partitioning and global prior injection of a learnable prototype library, combined with residual fusion, the stability and robustness of features are realized.
It improves the detail integrity and overall structural consistency of the super-resolution reconstruction results, reduces computational complexity, and enhances the model's reconstruction robustness in lightweight scenarios.
Smart Images
Figure CN121981892B_ABST
Abstract
Citation Information
Patent Citations
Hyperspectral image super-resolution reconstruction method based on multi-scale cavity convolution guidance
CN120852162A
Image super-resolution reconstruction method based on channel perception aggregation Transform
CN120976024A