基于特征聚类的图像错误检测及掩盖算法Errordetectionand.PDF

第31 卷第12 期 通 信 学 报 Vol.31 No.12 2010 年12 月 Journal on Communications December 2010 基于特征聚类的图像错误检测及掩盖算法 李国波，陈钢，吴百锋 （复旦大学计算机科学技术学院，上海 200433 ） 摘 要：针对现有的大多数图像错误掩盖算法缺乏错误检测能力，在空间域和变换域提出了2 个基于特征聚类并 结合信息隐藏技术的错误检测和恢复的掩盖算法，将图像 DCT 变换的直流系数和低频系数按块聚类，把块聚类 号作为隐藏信息分别嵌入到像素和中频系数的最低位平面中，通过这些隐藏的块聚类号之间的一致性比较实现了 错误块的检测，再用与错误块具有相同类的正常块的中心点数据替代错误块完成错误掩盖。对比实验表明，该算 法在平衡隐藏容量、隐写图像质量、掩盖效果和顽健性之间的矛盾上具有优势。 关键词：多媒体通信；错误掩盖；特征聚类；数据隐藏 中图分类号：TN919.8 文献标识码：A 文章编号：1000-436X(2010)12-0001-11 Error detection and concealment based on characteristic clustering of image LI Guo-bo, CHEN Gang, WU Bai-feng (School of Computer Science, Fudan University, Shanghai 200433, China) Abstract: Most existing concealing algorithms couldn’t detect transmitting errors. Two error detecting and concealing algorithms was proposed based on characteristic clustering (CCEDC) and information hiding techniques in spatial do- main and transform domain respectively. The proposed algorithms firstly clustered the DCT direct coefficients and low-frequency coefficients of blocks, and then embedded the clustering numbers into the least significant bit (LSB) plane of the corresponding pixels or middle-frequency coefficients. Therefore, the errors could be detected by comparing the consistency among clustering numbers and could be recovered using the clustering centers. The comparison experiments show that proposed algorithms have an advantage over other congener meth