基于Gram矩阵和卷积神经网络的风格迁移算法  

Style Transfer Algorithm Based on Gram Matrix and Convolutional Neural Network

作　　者：余志凡 李昊 李登实[1] 胡曦 YU Zhifan;LI Hao;LI Dengshi;HU Xi(School of Mathematics and Computer Science,Jianghan University,Wuhan 430056,Hubei,China)

机构地区：[1]江汉大学数学与计算机科学学院,湖北武汉430056

出　　处：《江汉大学学报（自然科学版）》2020年第3期62-68,共7页Journal of Jianghan University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61701194);湖北省教育厅科研计划资助项目(B2018254);国家级大学生创新创业训练计划资助项目(201911072015)。

摘　　要：如何获得更好的图像风格迁移效果一直是图像处理领域中经典问题之一。针对传统方法存在表现欠佳并具有较强局限性的问题,讨论了卷积神经网络(convolutional neural network,CNN)不同层提取图像特征的特点,提出了一种基于Gram矩阵和CNN的图像风格迁移算法,并设计了多种艺术风格的迁移实验,最后通过实验验证所提出的算法可比传统算法在更短的时间更好地实现艺术图像的风格迁移,说明该算法在图像特征提取及图像风格迁移任务上更具优势。How to get a better effect for the image style transfer is one of the typical problems in the field of image processing.Aiming at the problems of poor performance and strong limitation in the traditional methods,this paper firstly discussed on the characteristics of image features extraction in the different layers of CNN(convolutional neural networks),then an image style transfer algorithm based on Gram matrix and CNN was proposed,and a variety of art style transfer experiments were designed.The experimental results showed that the proposed algorithm could better implement the style transfer of art images in a shorter time than traditional algorithms,and the experiments verified the advantages of the proposed algorithm in image feature extraction and style transfer tasks.

关 键 词：图像风格迁移 卷积神经网络 GRAM矩阵 扩张卷积 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]