| 摘要: |
| 图像流认证的丢包鲁棒性问题是图像认证研究的难题之一。现有的流级认证算法通常是在信源-信道分离编码条件下针对某种特定的图像编码方式而设计的，在抵抗丢包能力方面具有很大的局限性。针对上述问题，提出一种丢包鲁棒的图像认证优化模型，并在此基础上提出了在信源-信道联合编码条件下实现信源-认证-信道码率的跨层优化资源分配（Cross-Layer Optimization Resource Allocation，CLORA）方法。首先以可信图像的端到端质量和认证代价为优化目标，结合基于图认证和基于前向纠错码（Forward Error Correction，FEC）认证方法，建立认证优化模型（Authentication Optimization Model，AOM），将图像认证的抗丢包优化问题等价成构造最优认证图（Optimal Authentication Graph，OAG）。然后利用图像码流的编码相关性和认证相关性，给出了求解OAG问题的等价条件，并在低计算复杂度下给出了构造OAG图的两个原子操作。最后提出了基于CLORA框架的认证优化方法。由于AOM模型仅利用码流的编码相关性信息，因而可以适用于不同的图像编码算法。JPEG 2000码流的实验结果表明，在相同丢包率下本文算法的端到端可信质量比已有算法平均提高了1.6dB，能够获得更优的丢包鲁棒性和端到端率失真（Rate-Distortion，R-D）性能。 |
| 关键词: 图像认证 丢包鲁棒性 可信质量 认证代价 信源信道联合编码 |
| DOI：10.19363/j.cnki.cn10-1380/tn.2018.12.010 |
| Received:June 19, 2016Revised:September 02, 2016 |
| 基金项目:本课题得到国家自然科学基金课题（No.61303259，No.U1536105）、国家科技支撑计划课题（No.2014BAH41B01）、国家重点研发计划课题（No.2016YFB0801003）、中国科学院战略性先导科技专项课题（No.XDA06030600）资助。 |
|
| An Image Stream Authentication Scheme Based on Cross-Layer Optimization Resource Allocation |
| YI Xiaowei,MA Hengtai,ZHAO Xianfeng,YU Haibo,ZHENG Changwen |
| State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China;Institute of Software Chinese Academy of Sciences, Beijing 100190, China |
| Abstract: |
| Packet-loss robustness of image stream authentication is one of the difficulties in image authentication research. Current stream-level authentication algorithms are designed for certain image coding under source-channel separate coding, so there are very limited on resistance of packet loss. In this paper, an image authentication optimization model with packet-loss robustness and a cross-layer optimization resource allocation (CLORA) of source-authentication-channel bitrate under joint source-channel coding are consecutively proposed to solve the above-mentioned problem. Firstly, in order to achieve the optimum end-to-end quality of the authentic image and authentication overhead, an authentication optimization model (AOM) is proposed based on combining two kinds of authentication methods, which include graph-based method and forward error correction (FEC) based method. Therefore, under the AOM, the optimization problem of packet-loss robustness in image authentication is equivalent to constructing the optimal authentication graph (OAG). Secondly, equivalent condition of the OAG is solved by using the coding-dependence of the image streams and the authentication relationship. And then according to the equivalent condition, two elementary operations to construct OAG are given with a very low computational complexity. Finally, an optimized authentication method is proposed by using the CLORA framework. Because the coding-dependence of the codestreams only is utilized in the AOM, the proposed AOM is suited to all kinds of image coding algorithms. Experimental results of JPEG 2000 streams demonstrate that the end-to-end authentic quality of the propose scheme increases by 1.6 dB than the existing authentication schemes with the same packet-loss rates. Therefore, our scheme obtains better packet-loss robustness and better end-to-end rate-distortion (R-D) performance. |
| Key words: Image authentication packet-loss robustness authentic quality authentication overhead joint source-channel coding |
