A Novel Digital Watermarking Algorithm Based On Scale-invariant Feature Regions In Non-subsampled Contourlet Transform Domain

With the development of the theory of multi-scale geometric analysis, it is widely used in the field of image processing. As a typical multi-scale geometric analysis transform, the Non-Subsampled Contourlet Transform (NSCT) provides a new way of thinking for watermarking with its characteristics of multi-resolution, multi-directional, translation invariance, locality and others. In this paper, by researching on watermark’s embedding position, a digital watermarking algorithm based on non-subsampled contourlet transform domain was proposed. Considering a stable scale-invariant feature transform (SIFT) features region, a digital watermarking algorithm with scale-invariant feature transform features concentric annular region based on non-subsampled contourlet transform domain was proposed. Experimental results show that the algorithm was very good trade-off of invisibility and robustness, which got a very good property, on the same time, it can effectively resist common image processing and geometric attacks and combo attacks. The main contents of this paper have the following aspects:(1) Researching on non-subsampled contourlet transform. Focusing on the characteristics of Non-Subsampled Contourlet Transform on the basic of the multi-scale geometric analysis, which makes good foundation for the following study of the non-subsampled transform applied to the digital watermarking.(2) Researching on watermarking embedding position based on non-subsampled contourlet transform domain. The characteristics of Non-subsampled contourlet transform offer a lot of choices in terms of the digital watermark embedding position. So this paper studied on the watermarking embedding position based on non-subsampled contourlet transform domain, at the same time comparing the performance on different watermark embedding transform domain from different scales and different frequencies, at last appropriate watermark embedding positions were obtained, which were settled on the largest energy sub-band after two layer decomposition. And the watermarking algorithm based on the selected watermarking embedding position obtained very good performance.(3) Combined with non-subsampled contourlet transform domain and scale-invariant feature transform. Most previous digital watermarking algorithms did not adequately consider the characteristics of the image content itself. So, the study on scale-invariant feature transform was applied to the above research of non-subsampled transform domain was presented in this paper, then digital watermarking algorithm based on non-subsampled transform domain and scale-invariant feature transform characteristic regional was proposed. A concentric circular pattern was designed to embed watermark information repeatedly and watermark embedding method made use of the quantization of parity, parity detector was designed to extract the watermark. Finally, simulation experiments results show that the algorithm has good balance between robustness and invisibility which were originally contradictory, at the same time, compared to other literatures the algorithm obtained better performance.