Article Preview
TopIntroduction
Digital watermarking for medical image has been studied extensively, and most of the proposed schemes have achieved the goal of the copyright protection (Badran, 2009; Bouslimi, Coatrieux, & Roux, 2012; Danyali, 2011; Fallahpour, Megias, & Ghanbari, 2011; Irany, Guo, & Hatzinakos, 2011; G. Lin, Tiegang, Guorui, Yanjun, & Li, 2012; Rahimi & Rabbani, 2011; Velumani & Seenivasagam, 2010). However, in practical situation, a slight tampering without causing significant visual distortion may totally destroy the diagnostic value of the medical image. So, the watermarking scheme with the purpose of tampering detection and recovery is preferred. At present, people have given many watermarking schemes which achieved the tamper detection and recovery for natural images (Bohra & Farooq, 2009; Bravo-Solorio & Nandi, 2011; Chamlawi & Khan, 2010; Chuang & Hu, 2011; Iliyasu, Le, Dong, & Hirota, 2012; Kim, Lee, Lee, Oh, & Lee, 2011; Lee & Rhee, 2006; C. Li, Wang, Ma, & Zhang, 2012), but some of them can’t be directly used for medical image, because medical application requires high visual quality of image after the embedding of the watermark. To enhance image visual quality, some frequency domain watermarking schemes with tamper detection and recovery have been proposed (Bohra & Farooq, 2009; Irany et al., 2011; C. Li et al., 2012; Velumani & Seenivasagam, 2010), as the frequency transform distributes the energy of the watermark to the whole image, the distortion caused by the watermark embedding was reduced, but it also has disadvantages such as low embedding capacity and high complexity of computation.
On the other hand, embedding watermark in the spatial domain supports relative higher embedding capacity and decreases the complexity of the computation, however, spatial domain embedding tamper detection methods usually have security flaws because the watermark is easier to be detected and decoded than in frequency domain. Since the security issue is very important in the evaluation of a watermarking system, it must be seriously considered when the watermarking scheme is carried out in spatial domain.
Recently Xiao et al. proposed a Piecewise Linear Chaotic Map (PWLCM) based watermarking scheme (Hong & Xie-Ting, 1997), which improved LSB watermarking scheme with the high security and high precision tampering detection (Xiao & Shih, 2012). Based on the scheme proposed by Lin et al.(P. L. Lin, Hsieh, & Huang, 2005), Xiao(2012) introduce the PWLCM to solve the security flaws of Lin's scheme(Chang, Fan, & Tai, 2008). The theoretical analysis and experimental results show the application of PWLCM can meet the need of security. However, as Xiao(2012)'s scheme is still a block-based embedding and detection scheme embedding the block's average value in another blocks, it is vulnerable to VQ attacks (Holliman & Memon, 2000). To fix this problem, Xiao(2012) introduced an extra level of detection in the procedure of tamper detection, which increased the complexity of computation. Besides this, the block-wise precision of tamper detection can not meet the need of medical usage. A pixel-wise precision method would be preferred even the recovery quality is not as good as the block-based scheme. This is because a precise localization can provide the user an exact reference of the tampered region. For medical usage, this is critical for avoiding misdiagnosis.