期刊
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
卷 44, 期 2, 页码 955-968出版社
IEEE COMPUTER SOC
DOI: 10.1109/TPAMI.2020.3014629
关键词
Feature extraction; Neural networks; Semantics; Image edge detection; Fuses; Task analysis; Machine learning; Defocus blur detection; multi-scale features; feature fusing; channel attention
资金
- National Natural Science Foundation of China [61701451, 61773392, 61901205, U1711266, 41925007]
- Opening Fund of Key Laboratory of Geological Survey and Evaluation of Ministry of Education [GLAB2020ZR18]
- Fundamental Research Funds for the Central Universities
In this paper, a deep neural network called DeFusionNet is proposed for defocus blur detection. DeFusionNet addresses challenges such as background clutter interference, scale sensitivity, and missing boundary details of blur regions by recurrently fusing and refining multi-scale deep features. The network incorporates feature adaptation and channel attention modules to narrow the gap between low-level and high-level features and select discriminative features for feature refinement. Extensive experiments on multiple datasets demonstrate the efficacy and efficiency of DeFusionNet.
Albeit great success has been achieved in image defocus blur detection, there are still several unsolved challenges, e.g., interference of background clutter, scale sensitivity and missing boundary details of blur regions. To deal with these issues, we propose a deep neural network which recurrently fuses and refines multi-scale deep features (DeFusionNet) for defocus blur detection. We first fuse the features from different layers of FCN as shallow features and semantic features, respectively. Then, the fused shallow features are propagated to deep layers for refining the details of detected defocus blur regions, and the fused semantic features are propagated to shallow layers to assist in better locating blur regions. The fusion and refinement are carried out recurrently. In order to narrow the gap between low-level and high-level features, we embed a feature adaptation module before feature propagating to exploit the complementary information as well as reduce the contradictory response of different feature layers. Since different feature channels are with different extents of discrimination for detecting blur regions, we design a channel attention module to select discriminative features for feature refinement. Finally, the output of each layer at last recurrent step are fused to obtain the final result. We collect a new dataset consists of various challenging images and their pixel-wise annotations for promoting further study. Extensive experiments on two commonly used datasets and our newly collected one are conducted to demonstrate both the efficacy and efficiency of DeFusionNet.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据