基于LFRs的大容量抗几何攻击水印算法*

如何有效抵抗几何攻击是目前水印技术研究的难点之一，且已有抗几何攻击水印算法的嵌入容量有限。提出了一种利用尺度不变特征变换（SIFT）特征点确定的局部特征区域（LFRs）嵌入水印的大容量抗几何攻击的水印算法。算法首先从图像的非抽样的Contourlet域（NSCT）低频分量中提取出SIFT中等尺度的稳定特征点，并通过最小生成树聚类算法的选择策略获得一组分布均匀且彼此独立的圆形特征区域；然后对每个圆形区域进行等角度的扇形分割和等面积的同心圆环分割，划分成等面积的子块；最后采用奇偶量化将经过混沌加密的水印嵌入到

High capacity image watermarking scheme robust to geometric attacks based on local feature regions

Image watermarking resistant to geometric attacks is the hotspot and challenging task in the state of the research on watermarking, and the capacity of these watermarking schemes existed usually is limited. This article presented an image watermarking scheme robust to geometric attacks based on local feature regions (LFRs), which were selected by scale-invariant feature transform (SIFT) keypoints. Firstly,it extracted the stable keypoints in medium scale from the non-subsampled Contourlet transform (NSCT) low-frequency of the host image by the SIFT algorithm, and got a set of widely-distributed and non-overlapped circle feature regions via selection criterion based on the minimum spanning tree clustering algorithm. Then every circle region was partitioned to sub-blocks with the same area in a sector pattern with the same angle and a homocentric cirque pattern with the same area. Finally, using odd-even quantization, the watermark, which had been encrypted by chaos system, could be embedded into the NSCT low-frequency of the sub-blocks. During the watermarking detection, the synchronous local regions could be gotten by the SIFT match algorithm, then could extract the watermark blindly. Experimental results show that the capacity of the watermark can be significantly increased, and the proposed scheme is robust to a wide range of image processing operations such as adding noise, filtering and compression and all kinds of geometric attacks.