一种自上而下属性不敏感的多尺度沙漏去雨网络及方法
By employing a top-down, attribute-insensitive, multi-scale hourglass rain removal network, and utilizing spatial and feature processing modules with decreasing resolution, the problem of increased computational cost and inference time in existing technologies is solved, achieving efficient removal of rain streaks and raindrops while balancing network performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIAN INST OF OPTICS & PRECISION MECHANICS CHINESE ACAD OF SCI
- Filing Date
- 2023-10-20
- Publication Date
- 2026-07-17
AI Technical Summary
With the improvement of hardware computing power, existing rain removal methods have become increasingly complex in terms of network structure, leading to increased computational costs and inference time. They are also difficult to effectively remove rain streaks and raindrops, and lack universal adaptability, especially in the processing of rain layers with different properties.
A top-down, attribute-insensitive multi-scale hourglass rain removal network is adopted. Through low, medium, and high-level spatial resolutions with decreasing resolution, combined with feature scaling, feature extraction and reconstruction modules, and feature fusion and reconstruction modules, channel compression and expansion strategies are used to achieve effective separation of rain layer information.
While reducing computational costs and inference time, it can effectively remove rain streaks and raindrops from rainy images, achieving a good balance between network performance and inference speed, and is suitable for processing rain layers with different properties.
Smart Images
Figure CN117635468B_ABST