高位有效位概率算术解码的图像密写改进算法

摘要

提出一种基于MSBs(高位有效位)概率算术解码的DCT(离散余弦变换)域YASS(yet another steganographic scheme)密写改进算法.为了克服YASS算法因量化索引调制(QIM)嵌入特性造成的易攻击缺陷,在YASS消息嵌入中用原型信号MSBs概率分布约束的算术解码,使消息嵌入流的解码序列具有与原型信号一致的边缘分布,在保持YASS算法高随机性嵌入优点的同时,能使密写前后一阶分布直方图之间的差异最小化,有效保证了密写的安全性.实验结果表明,与YASS对比,该方法使系数分布直方图的平均变化量更小,其平均失真为6.3％,相比YASS 的平均失真6.8％,减少了0.5％,改进算法的抗隐写分析能力比后者更强,是一种有效的密写算法.%To overcome the attacking-prone drawbacks of quantization index modulation (QIM) in YASS (yet another steganographic scheme), an improved YASS algorithm based on MSBs (most significant bits) arithmetic decoding in DCT domain is proposed. First, a secret sequence is viewed as an arithmetic coding stream and recovered by arithmetic decoding with the MSBs probability of a cover signal distributed, thus the codec reset results in a same marginal distribution between decoding stream and the cover. Then, the decoded bits are reembedded as the convention. Combined with the highly random embedding of YASS, the proposed scheme has a higher security with the first order distribution historgram between the cover and the stego minimized. Experiments demonstrate that the proposed method outperforms YASS both in capability of anti-steg-analysis and the performance of coefficient distribution historgram, which is 6. 8% for YASS and 6. 3% for the improved one in average. It is an effective steganographic algorithm.