一种特征融合方法、图像去雾方法及装置
By using residual dense block processing and difference sampling addition fusion method, the problem of limited feature diversity in multi-scale feature fusion mode is solved, and the effective fusion of multi-scale features and the improvement of image dehazing performance are achieved.
CN115880192BActive Publication Date: 2026-07-17BEIJING ZITIAO NETWORK TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING ZITIAO NETWORK TECH CO LTD
- Filing Date
- 2021-09-27
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing multi-scale feature fusion methods limit the feature diversity in network architecture, affecting the learning ability of image dehazing.
Method used
The target features are processed using Residual Dense Blocks (RDB), and multi-scale features are fused by feature partitioning and difference sampling to generate the fusion result of the target features and the features to be fused.
Benefits of technology
During multi-scale feature fusion, the generation of new features is guaranteed, the feature diversity in the network architecture is enhanced, and the performance of image dehazing is improved.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115880192B_ABST
Abstract
本发明实施例提供了一种特征融合方法、图像去雾方法及装置，涉及图像处理技术领域。该方法包括：获取目标特征和至少一个待融合特征，所述目标特征和所述至少一个待融合特征分别为同一图像的不同空间尺度的特征；将所述目标特征划分为第一特征和第二特征；基于残差稠密块RDB对所述第一特征进行处理，获取第三特征；对所述第二特征和所述至少一个待融合特征进行融合，获取第四特征；合并所述第三特征和所述第四特征，生成所述目标特征和至少一个待融合特征的融合结果。本发明实施例用于解决现有技术中的多尺度特征融合方式会限制网络架构中的特征的多样性的问题。
Need to check novelty before this filing date? Find Prior Art