基于块匹配和校验格编码的大嵌入率图像隐藏方法

摘要

针对通过互联网传输秘密图像所需的安全性和隐蔽性, 结合多层校验格编码提出一种基于矢量量化的图像隐藏方法. 首先根据可重构原则, 利用载体图像的多层 MSB 构造候选匹配分块集合; 然后使用块匹配过程, 从该集合中为每个秘密图像分块寻找其最优匹配并记录索引值, 对于无最优匹配的分块, 使用 K 均值聚类从中选出部分代表分块作为其最优匹配, 并替换块匹配过程中未使用的候选匹配分块; 再使用 Huffman 编码压缩最优匹配分块的索引值、代表分块等数据; 最后使用多层校验格编码将压缩数据嵌入到载体图像的多层LSB. 实验结果表明, 与已有的4种图像隐藏方法相比, 该方法生成的载密图像的视觉质量更好, 同时提取图像的视觉质量更好; 可有效地抵抗3种常见的结构化隐写分析方法; 具有嵌入容量大、载密图像有良好的视觉质量、提取图像和秘密图像的差异小等优点.%Aiming at the security and secrecy of transmitting secret images through the Internet, combined with the multi-layered syndrome-trellis codes, this paper proposed a vector-quantization-based image-hiding method. First of all, according to the reconfigurable principle, multi-layered most significant bits of the cover image were used to generate a set of candidate blocks. Then a block matching process was used to find the best-matching block for each block of secret image from the set and record the index values, and theK-means clustering was used to select some representative blocks from the not-well-matched blocks as the best-matching blocks for them to substitute the unused candidate blocks in the block matching process. Furthermore, the Huffman coding scheme was used to compress the index values of the best-matching blocks, representative blocks, etc. Finally, the encoded data were embedded in the multi-layered least significant bits of the cover image using the multi-layered syndrome-trellis codes. The experimental results show that, compared with four previous methods, the visual quality of the stego images obtained by the proposed method is better, and the visual quality of the extracted im-ages is also better; the proposed method can resist the detection of three common structural attacks; the proposed method exhibits the following advantages: a high embedding capacity; a good visual quality of the stego image; a small error between the extracted image and the secret image.