基于图像分割的矢量量化数字水印算法

o improve codebook quality in the process of vector quantization, this paper proposes a codebook generation algorithm which is based on image segmentation. Image regions with different characteristics are quantized separately. A preprocessing step is introduced with respect to the position of watermark embedding and the way of watermark extracting. In the watermark embedding process, the embedding position is chosen as the nearest distance in the pairing-table. In watermark extracting process, the codebook is checked for outliers firstly. If this is the case, the average value of the surrounding values is used instead. Experimental results show that the proposed algorithm not only keeps most of the details of the original image and the codebook is in higher quality, but also is robust to common image processing operations. In addition, the proposed algorithm has some anti-attack capability for tampering with codebook and index values.