A Rotation, Scaling, and Cropping Invariant Second Generation Watermarking Scheme Based on Hough Transform

* sponsored by Chinese National Science Foundation(Grant No.60172331), contact: [email protected]; phone 86 755 6536145; fax: 86 755 6536198; http://www.szu.edu.cn; Faculty of Information Engineering, Shenzhen University, Shenzhen, Guangdong Province, China, 518060; Abstract In this paper, a scheme which follows the second generation watermarking(2GW) paradigm is proposed. The goal of this proposed scheme is basically to increase the robustness against geometric attacks. The host image is decomposed with the wavelet packet. The bit stream of binary watermark is coded into several patterns with salient feature. The circular feature is used in this paper, because: (a) the computational complexity of the method grows rapidly with more complex shapes. (b) the circle is rotation-invariant and partially scale-invariant. (c) the circular feature can be detected by Hough transform effectively. These patterns are embedded into wavelet packet coefficients according to human perceptual characteristics. A new HVS mask based on wavelet transform is proposed with consideration of local texture characteristics. The introduction of HVS characteristics boosts the performance of the whole scheme. Then, the watermarked image is generated by WP reconstruction. For the detection of watermark, the geometric distortion is calibrated for the contaminated watermarked image, then it is decomposed with the same WP used in embedding procedure. Hough transform is used to detect the circular features in the WP Coefficients. This scheme has the following characteristic s: (a) robustness against the common geometric attacks(rotation, scaling, cropping, and etc) is improved significantly. (b) human perceptual characteristics is taken into consideration, so the tradeoff between invisibility and robustness is improved. Results of extensive experiments indicate that this proposed scheme is significantly effective in resisting various attacks such as rotation, scaling, JPEG compression, adding noise, etc.