Quantization parameter based blind watermarking schemes are very popular. These schemes work in spatial domain as well as in various transform domains such as wavelet, discrete cosine, and real oriented wavelet. These schemes utilize one bit (binary) image as a watermark. However, in general, watermarks are any n bit image, where n> 1. Therefore, these schemes are not suitable for the watermarks with n> 1. In this paper, we extend the existing work on blind watermarking schemes for n bits watermark images. The 'division algorithm' is used as a key component to develop the extended schemes. The proposed scheme is implemented in spatial domain. Experiments are done on one bit, two bits and eight bits watermark images. The results are investigated qualitatively by visualization, and measured quantitatively by peak-signal-to-noise-ratio (PSNR) and normalized correlation coefficient (NC). The results of the existing and proposed watermarking schemes are same for n=1. The results of the proposed watermarking scheme are reliable for n> 1. A formula is proven for lower bound on PSNR, which is a key finding in the watermarking community. The formula is verified experimentally.
展开▼
