基于神经网络与注意力的任意图像风格迁移研究综述  

Overview of Arbitrary Image Style Transfer Based on Neural Networks and Attention

作　　者：王树声 李文书[1] WANG Shusheng;LI Wenshu(School of Computer Science and Technology,Zhejiang Sci-Tech University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学计算机科学与技术学院,浙江杭州310018

出　　处：《软件工程》2025年第2期27-31,共5页Software Engineering

摘　　要：随着神经网络的广泛应用,尤其是注意力机制的引入,风格迁移研究取得了显著进展。文章对基于卷积神经网络、注意力机制的图像风格迁移进行了综述。首先,分析了风格迁移的基本原理和传统方法,详细介绍了基于深度学习的风格迁移算法,尤其聚焦于那些通过引入注意力机制来强化模型风格表现与内容保持能力的创新方法。其次,通过比较不同算法的性能,探讨了现有方法在局部内容保留和风格控制精度方面的优点与缺点。最后,分析了任意图像风格迁移领域的发展趋势和潜在的研究方向。With the widespread application of neural networks,particularly the introduction of attention mechanisms,significant progress has been made in style transfer research.A comprehensive review has been conducted on image style transfer based on convolutional neural networks and attention mechanisms.Firstly,the fundamental principles and traditional methods of style transfer are analyzed,with a detailed introduction to style transfer algorithms based on deep learning,especially those that incorporate attention mechanisms to enhance the stylization level and content retention of the models.By comparing the performance of different algorithms,the strengths and weaknesses of existing methods in terms of local content retention and style control precision are discussed.Finally,the development trends and potential research directions in the field of arbitrary image style transfer are analyzed.

关 键 词：图像风格迁移 深度神经网络 注意力机制 内容保留 风格化控制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]