TY - GEN
T1 - Edge Feature Fusion-Based Salient Object Detection Network
AU - Sun, Yifan
AU - Ren, Xuemei
AU - Jiang, Haiyan
AU - Zheng, Dongdong
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.
PY - 2026
Y1 - 2026
N2 - Recently, Salient Object Detection (SOD) is widely applied in numerous fields of computer vision, such as image recognition and segmentation. The existing SOD algorithms still have the problem of insufficient extraction of edge information. To address this issue, this paper proposes a salient object detection model utilizing edge feature fusion. Incorporating the feature extraction network, this model combines the attention mechanism and the edge feature extraction module, further improving the ability of the model to represent edge features. The Laplacian pyramid extracts salient edges of the image, connecting multiple feature extraction sub-modules through the side path to achieve the complementarity and fusion of salient target features and edge features. The channel and spatial attention modules form a series structure to conduct adaptive feature refinement for the fused features. The performance of the algorithm has been proved through experiments.
AB - Recently, Salient Object Detection (SOD) is widely applied in numerous fields of computer vision, such as image recognition and segmentation. The existing SOD algorithms still have the problem of insufficient extraction of edge information. To address this issue, this paper proposes a salient object detection model utilizing edge feature fusion. Incorporating the feature extraction network, this model combines the attention mechanism and the edge feature extraction module, further improving the ability of the model to represent edge features. The Laplacian pyramid extracts salient edges of the image, connecting multiple feature extraction sub-modules through the side path to achieve the complementarity and fusion of salient target features and edge features. The channel and spatial attention modules form a series structure to conduct adaptive feature refinement for the fused features. The performance of the algorithm has been proved through experiments.
KW - Deep learning
KW - Edge detection
KW - Salient object detection
UR - https://www.scopus.com/pages/publications/105040369588
U2 - 10.1007/978-981-95-6557-3_30
DO - 10.1007/978-981-95-6557-3_30
M3 - Conference contribution
AN - SCOPUS:105040369588
SN - 9789819565566
T3 - Lecture Notes in Electrical Engineering
SP - 307
EP - 318
BT - Proceedings of 2025 Chinese Intelligent Systems Conference - Volume 2
A2 - Jia, Yingmin
A2 - Zhang, Weicun
A2 - Fu, Yongling
A2 - Liu, Yang
PB - Springer Science and Business Media Deutschland GmbH
T2 - 21st Chinese Intelligent Systems Conference, CISC 2025
Y2 - 25 October 2025 through 26 October 2025
ER -