基于深度学习的轻量级实时图像分割方法研究  

Research on lightweight real-time image segmentation methods based on deep learning

作　　者：李建锋[1,2] 熊明强 陈园琼[1,2,3] 王宗达 向涛 孙培玮 LI Jianfeng;XIONG Mingqiang;CHEN Yuanqiong;WANG Zongda;XIANG Tao;SUN Peiwei(School of Communication and Electronic Engineering,Jishou University,Jishou 416000,China;School of Computer Science and Engineering,Jishou University,Jishou 416000,China;School of Computer,Central South University,Changsha 410083,China)

机构地区：[1]吉首大学通信与电子工程学院,湖南吉首416000 [2]吉首大学计算机科学与工程学院,湖南吉首416000 [3]中南大学计算机学院,湖南长沙410083

出　　处：《通信学报》2025年第2期176-190,共15页Journal on Communications

基　　金：国家自然科学基金资助项目(No.61962023)。

摘　　要：针对深度学习在各领域应用中因模型复杂度提升而引发的计算与存储负担,尤其在图像分割任务中面临的算法复杂性、实时响应不足及高内存占用问题,提出了一种轻量级且高效的分割网络架构——多尺度叠加融合网络(MSFNet)。MSFNet设计了一个双分支多尺度边界融合模块,该模块通过融合不同尺度的特征信息与边界细节,有效提升了图像分割精度,同时显著减少了模型参数量。实验结果表明,MSFNet在3个公开数据集上表现优异,其模型参数量仅为0.6×10^(6),在RTX 3070 GPU上处理大小为800像素×800像素的图像仅需12 ms,显著提升了分割任务的执行效率和资源利用率。因此,该模型特别适合应用于资源有限的边缘设备或移动设备中,为实时图像分割应用提供了有力的技术支撑。In response to the computational and storage burdens caused by the increasing model complexity in deep learning applications,especially in image segmentation tasks where algorithmic complexity,insufficient real-time responsiveness,and high memory usage were prevalent,a lightweight and efficient segmentation network architecture——multiscale superposition fusion network(MSFNet)was proposed.MSFNet featured a dual-branch multi-scale boundary fusion module,which effectively enhanced segmentation accuracy by integrating feature information and boundary details from different scales.At the same time,it significantly reduced the model parameter count.Experimental results show that MSFNet outperforms other models on three public datasets,with a model size of only 0.6×10^(6) parameters.On the RTX 3070 GPU,it processes 800×800 pixels images in just 12 ms,significantly improving the execution efficiency and resource utilization of segmentation tasks.Therefore,this model is particularly well-suited for deployment on resourceconstrained edge or mobile devices,providing a favorable technical foundation for real-time image segmentation applications.

关 键 词：图像分割 轻量级实时网络 双分支多尺度边界融合模块 

分 类 号：TN92[电子电信—通信与信息系统]