媒体鲁棒哈希函数的分析模型及性能极限研究

项目名称： 媒体鲁棒哈希函数的分析模型及性能极限研究

项目编号： No.61202164

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 李岳楠

作者单位： 天津大学

项目金额： 23万元

中文摘要： 鲁棒哈希函数是从图像和视频等媒体信息到其内容摘要的单向映射，由于可简洁地刻画媒体的感知内容，因而被广泛用于解决包括版权管理在内的各类媒体内容识别问题。在版权管理等应用中，明确鲁棒哈希函数的识别能力极限是算法选择过程中面临的首要问题。但由于缺乏合适的分析模型，现有研究还尚未能揭示鲁棒哈希函数的这一性能极限，由此造成了算法选择及算法性能评价方面的局限性。本项目致力于在信息论的框架下对鲁棒哈希函数的基础问题进行研究，力图通过理论建模和模型分析来明确其性能极限。项目从建立鲁棒哈希函数的分析模型为切入点，将基于鲁棒哈希函数的媒体识别问题转化为在带噪声信道上的信息传输问题；进而以信道编码理论为基础，从信道容量的角度研究鲁棒哈希函数识别能力的理论极限。项目旨在构建鲁棒哈希函数的分析模型并探寻其理论基础，以开拓鲁棒哈希函数基础问题研究的新思路；揭示鲁棒哈希函数的性能极限，为算法选择及评价提供指导。

中文关键词： 内容识别；鲁棒哈希函数；分析模型；性能极限；

英文摘要： Robust hash function (RHF) is a one-way mapping from media data (e.g., digital image and video) to its content digest. Owing to its capability of capturing the perceptual essence of media data, RHF has been extensively applied in various media identification applications, such as copyright management. In practical applications, understanding the performance bound of a given RHF in content identification is one of the primary concerns in algorithm selections. However, due to the lack of the analytical model, the current research findings are not able to reveal this performance bound of RHF, which leads to the limitations in algorithm selections and performance assessment. Motivated by this fact, the proposed project aims at investigating the basic principles of RHF within the framework of information theory. In particular, we focus on estimating the performance bound of RHF in content identification through modeling and analysis. By establishing the analytical model of RHF, we transform the RHF based media identification into the problem of information transmission over the noisy channel. Consequently, in light of the channel coding theorem，the performance bound of RHF is studied from the perspective of channel capacity. To sum up, the significance of the proposed research is twofold. First, establishing the anal

英文关键词： Content identification；Robust hash function；Analytical model；Performance bound；