A blind and robust video watermarking method based on hybrid 3-D transform

Digital images and videos can be copied, reproduced, and distributed with the same quality as the original ones, and this violates the copyright of original producers and the distributors. As a result, embedding information about the original producer and distributor in digital images and video attracted great attention for digital right management. Watermarking provides the facility to embed the required information in images and videos. Robust watermarking is used for embedding authentication information and hence should be robust against various attacks. On the other hand, in fragile watermarking the embedded data should be destroyed by any alteration in the watermarked image or video. Reversible digital watermarking techniques are proposed for lossless restoration of the original image from the watermarked image. Watermarking is non-blind when a copy of the signature or other related information is required for extracting the signature from the watermarked image or video and is blind when the signature can be extracted from watermarked data without any other subsidiary information. Watermarking the signature in a group of successive frames of a video file by the use of 3-D transforms attracted attention because it makes the watermarked video more robust against attacks such as frame averaging and alteration, and watermarking by 3-D transform is employed as a solution to the problems caused by independent watermarking of signature in one or several frames.
Contourlet transform offers a high degree of directionality and anisotropy besides the multi-scale and time-frequency localization properties in wavelet transform. As a result, Contourlet transforms the representation of curved edges in the images with smoother contour and fewer coefficients compared to the wavelet transform. In this paper, a blind robust watermarking method based on a hybrid 3-D transform is proposed. The hybrid 3-D transform is derived by employing the 2-D Contourlet transform along with the 1-D wavelet transform. The signature will be watermarked in the low-frequency sub-band derived from the third level transform. To watermark the signature, we save a modified copy of the high energy coefficients of the even part in the odd part. For signature extraction, the watermarked region is partitioned into odd and even columns. The 3-level 3-D is applied to odd and even parts to transform coefficients. The high-energy sub-bands in odd and even parts are separated to extract the signature.
Experimental results indicate low degradation of the quality of the watermarked video, along with high robustness of the watermarked video against common attacks in comparison with other tested blind video watermarking methods.
A comparison of the proposed method with other methods indicates the superior performance of the proposed method in most of the attacks.