Hiding Watermarks in Multi-Regions of Cover Images using the Wavelet Transform

In this paper, we investigate the embedding of a copyright watermark in four non-overlapping regions of a standard image after partitioning the image into four equal parts (sub images). The entropy of the LH and HL detail components in each sub image is calculated and the maximum of the two components’ (LH and HL) is chosen for embedding the watermark. The HH and LL components are not affected. The watermark is extracted using the correlation coefficient. The technique was tested using three standard images and the experimental results indicate that while a little visibility is compromised, the robustness is better than simple embedding of a single watermark. Various geometric and filtering attacks were applied on the watermarked image and the watermark was recovered successfully. Perceptual transparency was measured with PSNR while robustness was measured using the correlation coefficient (p). It is desired to have a value close to 1 for the correlation coefficient for maximum robustness. The PSNR of the extracted watermark after attack was in the range of 57 to 65 while the correlation coefficient values range of the extracted watermark was between 0.61 and 0.91, which is a very acceptable range.