Wavelet-Domain Image Watermarking Based on Statistical Metrics

Abstract

A new wavelet-based watermarking technique is presented in this paper, in which watermark signals are selected to be some gray-scale logo image(s). Discrete wavelet transform (DWT) is used to decompose each original image first, whereas the selected watermark logo is decomposed into bit-planes at the same time. In the embedding process, the wavelet coefficients of an original image are divided into blocks. The energy and standard deviation of these blocks are computed and used to control the inserting process. Significant bits of the watermark logo are embedded first in order to have more protection. Original image is necessary in the recovering process, and the differences of the energies and standard deviations between individual blocks in the original image and the watermarked image are used to determine the embedded bit being 1 or 0. Experiment results show that our new method yields quite good visual quality in watermarked images, and is robust to typical signal processing attacks such as compression and filtering.