AIGC视觉内容生成与溯源研究进展  

Review on the progress of the AIGC visual contentgeneration and traceability

作　　者：刘安安 苏育挺 王岚君 李斌[2] 钱振兴 张卫明[4] 周琳娜 张新鹏 张勇东 黄继武[2] 俞能海[6] Liu Anan;Su Yuting;Wang Lanjun;Li Bin;Qian Zhenxing;Zhang Weiming;Zhou Linna;Zhang Xinpeng;Zhang Yongdong;Huang Jiwu;Yu Nenghai(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China;College of Electronics and Information Engineering,Shenzhen University,Shenzhen 518060,China;School of Computer Science,Fudan University,Shanghai 200438,China;School of Information Science and Technology,University of Science and Technology of China,Hefei 230026,China;School of Cyberspace Security,Beijing University of Posts and Telecommunications,Beijing 100876,China;School of Cyber Science and Technology,University of Science and Technology of China,Hefei 230027,China)

机构地区：[1]天津大学电气自动化与信息工程学院,天津300072 [2]深圳大学电子信息与工程学院,深圳518060 [3]复旦大学计算机科学技术学院,上海200438 [4]中国科学技术大学信息科学技术学院,合肥230026 [5]北京邮电大学网络空间安全学院,北京100876 [6]中国科学技术大学网络空间安全学院,合肥230027

出　　处：《中国图象图形学报》2024年第6期1535-1554,共20页Journal of Image and Graphics

基　　金：国家自然科学基金项目(U21B2024,U20B2047,U2336206,U20B2051,U23B2022,62371330,62202329,62172053)。

摘　　要：随着数字媒体与创意产业的快速发展,人工智能生成内容(artificial intelligence generated content, AIGC)技术以其在视觉内容生成中的创新应用而逐渐受到关注。本文旨在围绕AIGC视觉内容生成与溯源研究进展深入研讨。首先,针对图像生成技术进行探讨,从基于生成式对抗网络的传统方法出发,系统地分析了基于生成式对抗网络、自回归模型和扩散概率模型的最新进展。接着,深入探讨可控图像生成技术,突出了通过布局、线稿等附加信息以及基于视觉参考的方法来为创作者提供精确控制的技术现状。随着图像生成技术的革新和应用,生成图像的安全性问题逐渐浮现。而预先审核和过滤的技术手段已难以满足实际需求,故亟需实现生成内容的溯源来进行监管。因此,本文进而对生成图像溯源技术进行研讨,并聚焦水印技术在确保生成内容可靠性和安全性方面的应用。依据水印嵌入的流程节点,首先将现有的水印相关的生成图像溯源方法归为无水印嵌入的生成图像溯源、水印前置嵌入的生成图像溯源、水印后置嵌入的生成图像溯源以及联合生成的生成图像溯源并进行详细分析,然后介绍针对生成图像的水印攻击研究现状,最后对生成图像溯源技术进行总结和展望。鉴于视觉内容生成在质量和安全上的挑战,旨在为研究者提供一个视觉内容生成与溯源的系统研究视角,以促进数字媒体创作环境的安全与可信,并引导未来相关技术的发展方向。In the contemporary digital era,which is characterized by rapid technological advancements,multimedia con⁃tent creation,particularly in visual content generation,has become an integral part of modern societal development.Theexponential growth of digital media and the creative industry has attracted attention to artificial intelligence generated con⁃tent(AIGC)technology.The groundbreaking applications of AIGC in visual content generation not only have equippedmultimedia creators with novel tools and capabilities but also have delivered substantial benefits across diverse domains,which span from the realms of cinema and gaming to the immersive landscapes of virtual reality.This review comprehensiveintroduces the profound advancements within AIGC technology.Our particular emphasis is on the domain of visual contentgeneration and its critical facet of traceability.Initially,our discussions trace the evolutionary path of image generationtechnology,from its inception within generative adversarial networks(GANs)to the latest advancements in Transformerauto-regressive models and diffusion probability models.This progression unveils a remarkable leap in the quality and capa⁃bility of image generation,which underscores the rapid evolution of this field.This evolution has transitioned from itsnascent stages to an era characterized by explosive growth.First,we delve into the development of GANs,encompassingtheir evolution from text-conditioned methods to sophisticated techniques for style control and the development of largescale models.This type of technology pioneered the text-to-image generation.GANs can further improve their performanceby expanding network parameters and dataset size due to their strong scalability.Furthermore,we explore the emergence ofTransformer-based auto-regressive models,such as DALL·E and CogView,which have heralded a new epoch in thedomain of image generation.The basic strategy of autoregressive models is to first use the Transformer structure to predictthe feature sequence of images based

关 键 词：人工智能内容生成(AIGC) 视觉内容生成 可控图像生成 生成内容安全 生成图像溯源 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构] TP391.41[自动化与计算机技术—计算机科学与技术]