关键词: 不可感知性, 鲁棒性, Radon变换, 虚警率, 漏检率

Abstract: Most existing digital watermarking algorithms exceedingly focus on the imperceptibility and robustness of the watermarked image and geometric transform watermarked image.However，the analysis of false probability of watermarked image detection is insufficient.If a digital watermarking algorithm has a higher false probability in watermark detection，it can not solve copyrights problem efficiently although this algorithm has a good robustness.On the contrary，it may cause a new copyright dissension.For above reasons，this paper proposes a digital watermarking image resisting to geometric attacks based on radon invariant moments which can decrease the false probability of watermarked image.This algorithm uses the radon transform to construct invariant moments.Firstly，the characteristic of wavelet transformation is used to extract low frequency image.Secondly，this paper makes use of image scrambling to eliminate the corelation among image pixels，and uses zero pixels value model background image to replace image scale normalization，and then the low frequency image has been scrambled to zero pixels value model background image.Finally，the radon invariant moments are used to design watermark and detect watermark.The experiments demonstrate the proposed method is more robust to RST transforms，also it is more robust to adding noise，filtering and JPEG compression attacks.So the proposed method is a simple，useful and credible algorithm.

Key words: imperceptibility, robustness, Radon transform, false alarm probability, omission ratio