Evaluation of scanned image tolerance by using spread spectrum digital watermarking with regression analysis

摘要

In order to detect watermarks from a scanned image, we investigated color change between the stego-image and the scanned one. The color distribution of the scanned image varies depending on the distribution of the stego-image and the conditions of both printer and scanner. By this variation, it is hard to extract the watermarks from the scanned image. To improve robustness of the watermarks, we introduce the spread spectrum watermarking method in the embedder and decoder. The results of correlation between the stego-image and the scanned one by the regression method show that the mapping can be approximated by a regression line. Therefore, the color of the scanned image was corrected by using the regression line. As a result, the bit error rate for the watermarks obtained from the corrected image can be reduced.