基于结构引导的山水画分步风格迁移算法  

Structure-guided step-by-step style transfer algorithm for landscape paintings

作　　者：李弘宇 吴晓军[1] LI Hongyu;WU Xiaojun(School of Computer Science,Shaanxi Normal University,Xi'an 710119,China)

机构地区：[1]陕西师范大学计算机科学学院,西安710119

出　　处：《中国体视学与图像分析》2023年第2期212-218,共7页Chinese Journal of Stereology and Image Analysis

摘　　要：针对将数字照片转换为山水画问题,深入分析了山水画风格迁移的研究现状,讨论了当前存在结构信息关注不足等问题,并提出了一种基于结构引导的山水画分步风格迁移算法。该算法在网络架构和损失函数设计上进行改进,利用结构引导的山水画循环生成对抗网络EGS-CycleGAN(Edge-Guided Structural CycleGAN)来实现山水画的分步迁移。在中国山水画数据集上的实验结果表明,所提算法在图像质量和迁移效果方面表现优异,可以快速、高效地将数字图像转换为自然风格的山水画图像。This study focuses on the problem of transforming digital photographs into landscape paintings.With the current research status of style transfer in landscape painting analyzed,it addresses the insufficient attention given to structural information and proposes a step-by-step style transfer algorithm for landscape painting guided by underlying structures.The algorithm improves the network architecture and loss function design by utilizing an ECS-CycleGAN(Edge-Guided Structural CycleGAN)to achieve step-by-step migration of landscape paintings.Experimental results on a Chinese landscape painting dataset demonstrate the outstanding performance of the proposed algorithm in terms of image quality and transfer effects.It enables the rapid and efficient conversion of photographs into images with the style of landscape paintings.

关 键 词：结构引导 山水画风格迁移 分步迁移 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]