计算机科学与探索 ›› 2019, Vol. 13 ›› Issue (5): 721-741.DOI: 10.3778/j.issn.1673-9418.1810023

• 综述·探索 • 上一篇    下一篇

图像篡改检测感知哈希技术综述

杜  玲,陈  振+   

  1. 天津工业大学 计算机科学与技术学院,天津 300387
  • 出版日期:2019-05-01 发布日期:2019-05-08

Survey on Image Tamper Detection with Perceptual Hashing

DU Ling, CHEN Zhen+   

  1. School of Computer Science and Technology, Tianjin Polytechnic University, Tianjin 300387, China
  • Online:2019-05-01 Published:2019-05-08

摘要: 互联网的发展使得多媒体的真实性、完整性认证成为亟待解决的问题。感知哈希在理解图像内容基础上,通过简短的感知摘要来完成图像内容的识别和认证,为解决与多媒体认证相关的管理问题提供了一种更为便捷的方式。首先,总结了目前基于底层线索和基于学习的感知哈希图像篡改检测方法,并根据方法的不同特点进行了更为细致的分类。其次,介绍了常用的数据集,给出了三种算法性能评价指标,并在不同数据集上对最近的几种算法进行了定性和定量的比较分析。最后,对基于感知哈希的图像篡改检测的关键问题进行了总结,并对未来的发展趋势进行了展望。

关键词: 感知哈希, 篡改检测, 底层线索, 哈希学习

Abstract: The rapid development of internet techniques has brought many issues to multimedia data authenticity and integrity authentication. Perceptual Hashing can complete the identification and authentication of image content through a brief perceptual representation, which provides a more convenient way to solve the management problems related to multimedia data authentication. This paper first summarizes the methods of low-level cues and learning based perceptual Hashing algorithms for image tamper detection, and gives a more detailed classification based on the different characteristics. Next, this paper introduces the common used datasets for image tamper detection and three evaluation metrics. Four existing methods with qualitative and quantitative comparisons on different datasets are analyzed in the experiments. Finally, the key problems of image tamper detection based on perceptual Hashing are summarized, and the future development trend is prospected.

Key words: perceptual Hashing, tamper detection, low-level cues, learning based Hashing