Page 474 - 《软件学报》2026年第2期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
2026,37(2):953−968 [doi: 10.13328/j.cnki.jos.007443] [CSTR: 32375.14.jos.007443] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

*
基于边缘增强的宽解码器显著性目标检测方法

童 彪, 宋晓宁, 华 阳, 张文杰, 吴小俊

(江南大学 人工智能与计算机学院, 江苏 无锡 214100)
通信作者: 宋晓宁, E-mail: x.song@jiangnan.edu.cn

摘 要: 当前, 显著性目标检测技术正在迅速发展, 但仍然存在一些问题亟待解决. 大多数现有的显著性目标检测
方法在处理高分辨率图像任务时, 存在计算资源需求过高或者检测质量较差等问题. 其次, 许多现有算法采用的传
统卷积操作缺乏针对性, 无法有效增强边缘细节特征, 导致边缘分割模糊不清. 为了在降低算力消耗的同时提高物
体边缘分割质量, 并提升小尺度目标的检测性能, 提出了基于边缘增强的宽解码器显著性目标检测方法. 采用残差
网络和 Swin Transformer 组合结构作为特征编码器, 以降低算力消耗. 并且将传统卷积替换为差分卷积模块, 通过
多种不同类型的差分卷积并行使用, 从图像中提取了更加丰富的边缘信息. 设计了多尺度注意力模块, 对 4 层不同
尺度特征进行注意力计算, 以更好地关注不同大小的目标. 此外, 采用含有大卷积核的多级宽解码器, 对融合特征
进行长距离的上下文建模, 减少冗余信息, 进一步提升了网络的检测性能.
关键词: 显著性目标检测; 边缘增强; 差分卷积
中图法分类号: TP391

中文引用格式: 童彪, 宋晓宁, 华阳, 张文杰, 吴小俊. 基于边缘增强的宽解码器显著性目标检测方法. 软件学报, 2026, 37(2):
953–968. http://www.jos.org.cn/1000-9825/7443.htm
英文引用格式: Tong B, Song XN, Hua Y, Zhang WJ, Wu XJ. Salient Object Detection Method Based on Edge-enhanced Wide
Decoder. Ruan Jian Xue Bao/Journal of Software, 2026, 37(2): 953–968 (in Chinese). http://www.jos.org.cn/1000-9825/7443.htm

Salient Object Detection Method Based on Edge-enhanced Wide Decoder
TONG Biao, SONG Xiao-Ning, HUA Yang, ZHANG Wen-Jie, WU Xiao-Jun
(School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi 214100, China)

Abstract: Salient object detection is developing rapidly. However, several critical challenges remain. Most existing methods struggle with
high-resolution images due to either excessive computational demands or suboptimal detection quality. In addition, traditional convolutional
operations commonly used in current algorithms lack targeted enhancement, resulting in inadequate edge detail extraction and blurred
object boundaries. To address these limitations, this study proposes a salient object detection method based on an edge-enhanced wide
decoder, which improves edge segmentation accuracy and enhances small-scale object detection while reducing computational overhead. A
hybrid feature encoder combining a residual network and a Swin Transformer is employed to lower computational overhead. Traditional
convolutions are replaced with a differential convolution module, where multiple types of differential convolutions are executed in parallel
to extract richer edge information. A multi-scale attention module is incorporated to compute attention across four hierarchical feature
layers, enabling better focus on objects of varying sizes. In addition, a multilevel wide decoder with large convolutional kernels is utilized
to conduct long-range contextual modeling of fused features, effectively reducing redundant information and further boosting detection
performance. Code will be released at https://github.com/wapitier/EEWDNet.
Key words: salient object detection; edge enhancement; difference convolution

* 基金项目: 国家重点研发计划 (2023YFF1105102, 2023YFF1105105); 国家社会科学基金 (21&ZD166); 国家自然科学基金 (61876072);
江苏省自然科学基金 (BK20221535)
收稿时间: 2024-04-14; 修改时间: 2024-06-21, 2025-02-20; 采用时间: 2025-04-04; jos 在线出版时间: 2025-11-13
CNKI 网络首发时间: 2025-11-14