第47卷第2期红外与激光工程2018年2月Vol.47No.2Infrared and Lar Engineering Feb.2018笠特约专栏笠
基于对抗生成网络的纹理合成方法
余思泉1,2,韩志2,唐延东1,2,吴成东1
参观函(1.东北大学信息科学与工程学院,辽宁沈阳110000;
2.中国科学院沈阳自动化研究所机器人学国家重点实验室,辽宁沈阳110000)
摘要:纹理合成是计算机图形学、计算机视觉和图像处理领域的研究热点之一。传统的纹理合成方法往往通过提取有效的特征样式或统计量并在该特征信息的约束下生成随机图像来实现。对抗生成网络作为一种较新的深度网络形式,通过生成器和判别器的对抗训练能够随机生成与观测数据具有相同分布的新数据。鉴于此,提出了一种基于对抗生成网络的纹理合成方法。该算法的优点是不需要经过多次迭代就能够生成更真实纹理图像,且生成图像在视觉上与观测纹理图像一致的同时具有一定随机性。一系列针对随机纹理和结构性纹理的合成实验验证了该算法的有效性。
cpu超频
关键词:纹理合成;深度学习;生成模型;对抗生成网络销售部门
中图分类号:TP391文献标志码:A DOI:10.3788/IRLA201847.0203005
70年
Texture synthesis method bad on generative adversarial networks
Yu Siquan1,2,Han Zhi2,Tang Yandong1,2,Wu Chengdong1
(1.School of Information Science and Engineering,Northeastern University,Shenyang110000,China;
2.State Key Laboratory of Robotics,Shenyang Institute of Automation,Chine Academy of Sciences,Shenyang110000,China)
奇数是什么Abstract:Texture synthesis is a hot rearch topic in the fields of computer graphics,vision,and image processing.Traditional texture synthesis methods are generally achieved by extracting effective feature patterns or statistics and generating random images under the constraint of the feature information.
Generative adversarial networks(GANs)is a new type of deep network.It can randomly generate new data of the same distribution as the obrved data by training generator and discriminator in an adversarial learning mechanism.Inspired by this point,a texture synthesis method bad on GANs was propod.The advantage of the algorithm was that it could generate more realistic texture images without iteration;the generated images were visually consistent with the obrved texture im北京鸟巢简介
age and also had randomness.A ries of experiments for random texture and structured texture synthesis verify the effectiveness of the propod algorithm.
Key words:texture synthesis;deep learning;generative model;generative adversarial networks
收稿日期:2017-09-05;修订日期:2017-10-05
基金项目:国家自然科学基金(61773367,61303168);中国科学院青年创新促进会(2016183)
作者简介:余思泉(1988-),男,博士生,主要从事图像中层表达、深度学习等方面的研究。Email:yusiquan@sia
关于学习的四字词语导师简介:唐延东(1962-),男,研究员,博士生导师,博士,主要从事图像处理、目标识别方面的研究。Email:ytang@sia
长跪0203005-1