Abstract: Watermarking algorithms that use convolution neural networks have exhibited good robustness in studies of deep learning networks. However, after embedding watermark signals by convolution, the feature fusion efficiency of convolution is relatively low; this can easily lead to distortion in the embedded image. When distortion occurs in medical images, especially in diffusion tensor images (DTIs), the clinical value of the DTI is lost. To address this issue, a robust watermarking algorithm for DTIs implemented by fusing convolution with a Transformer is proposed to ensure the robustness of the watermark and the consistency of sampling distance, which enhances the quality of the reconstructed image of the watermarked DTIs after embedding the watermark signals. In the watermark-embedding network, T1-weighted (T1w) images are used as prior knowledge. The correlation between T1w images and the original DTI is proposed to calculate the most significant features from the T1w images by using the Transformer mechanism. The maximum of the correlation is used as the most significant feature weight to improve the quality of the reconstructed DTI. In the watermark extraction network, the most significant watermark features from the watermarked DTI are adequately learned by the Transformer to robustly extract the watermark signals from the watermark features. Experimental results show that the average peak signal-to-noise ratio of the watermarked DTI reaches 50.47 dB, the diffusion characteristics such as mean diffusivity and fractional anisotropy remain unchanged, and the main axis deflection angle α_AC is close to 1. Our proposed algorithm can effectively protect the copyright of the DTI and barely affects the clinical diagnosis.

弥散张量图像的鲁棒水印算法

[1]Ahmadi M, Norouzi A, Karimi N, et al., 2020. ReDMark: framework for residual diffusion watermarking based on deep networks. Exp Syst Appl, 146:113157.

[2]Amini M, Ahmad MO, Swamy MNS, 2018. A robust multibit multiplicative watermark decoder using a vector-based hidden Markov model in wavelet domain. IEEE Trans Circ Syst Video Technol, 28(2):402-413.

[3]Anand D, Niranjan UC, 1998. Watermarking medical images with patient information. Proc 20th Annual Int Conf of the IEEE Engineering in Medicine and Biology Society, p.703-706.

[4]Anctil-Robitaille B, Desrosiers C, Lombaert H, 2021. Manifold-aware CycleGAN for high-resolution structural-to-DTI synthesis. Proc Int MICCAI Workshop on Computational Diffusion MRI, p.213-224.

[5]Cintra RJ, Cooklev TV, 2009. Robust image watermarking using non-regular wavelets. Signal Image Video Process, 3(3):241-250.

[6]Dosovitskiy A, Beyer L, Kolesnikov A, et al., 2020. An image is worth 16×16 words: Transformers for image recognition at scale. https://arxiv.org/abs/2010.11929

[7]Feng CM, Yan Y, Fu H, et al., 2021. Task transformer network for joint MRI reconstruction and super-resolution. 24^th Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.307-317.

[8]He KM, Zhang XY, Ren SQ, et al., 2016. Deep residual learning for image recognition. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.770-778.

[9]Hu J, Shen L, Sun G, 2018. Squeeze-and-excitation networks. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.7132-7141.

[10]Huan WN, Li S, Qian ZX, et al., 2022. Exploring stable coefficients on joint sub-bands for robust video watermarking in DT CWT domain. IEEE Trans Circ Syst Video Technol, 32(4):1955-1965.

[11]Huang G, Liu Z, van der Maaten L, et al., 2017. Densely connected convolutional networks. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.2261-2269.

[12]Lai ZY, Qu XB, Liu YS, et al., 2016. Image reconstruction of compressed sensing MRI using graph-based redundant wavelet transform. Med Image Anal, 27:93-104.

[13]Le Bihan D, Mangin JF, Poupon C, et al., 2001. Diffusion tensor imaging: concepts and applications. J Magn Reson Imag, 13(4):534-546.

[14]Lee HK, Kim HJ, Kwon KR, et al., 2005. ROI medical image watermarking using DWT and bit-plane. Proc Asia-Pacific Conf on Communications, p.512-515.

[15]Li HY, Liang ZF, Zhang CY, et al., 2021. SuperDTI: ultrafast DTI and fiber tractography with deep learning. Magn Reson Med, 86(6):3334-3347.

[16]Liu XL, Lin CC, Yuan SM, 2018. Blind dual watermarking for color images’ authentication and copyright protection. IEEE Trans Circ Syst Video Technol, 28(5):1047-1055.

[17]Luo XY, Zhan RH, Chang HW, et al., 2020. Distortion agnostic deep watermarking. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.13545-13554.

[18]Nakarmi U, Wang Y, Lyu J, et al., 2017. A kernel-based low-rank (KLR) model for low-dimensional manifold recovery in highly accelerated dynamic MRI. IEEE Trans Med Imag, 36(11):2297-2307.

[19]Ni ZC, Shi YQ, Ansari N, et al., 2006. Reversible data hiding. IEEE Trans Circ Syst Video Technol, 16(3):354-362.

[20]Ravishankar S, Bresler Y, 2011. MR image reconstruction from highly undersampled k-space data by dictionary learning. IEEE Trans Med Imag, 30(5):1028-1041.

[21]Shin PJ, Larson PEZ, Ohliger MA, et al., 2013. Calibrationless parallel imaging reconstruction based on structured low-rank matrix completion. Magn Reson Med, 72(4):959-970.

[22]Singh AK, Dave M, Mohan A, 2015. Hybrid technique for robust and imperceptible multiple watermarking using medical images. Multim Tools Appl, 75(14):8381-8401.

[23]Stejskal EO, Tanner JE, 1965. Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient. J Chem Phys, 42(1):288-292.

[24]Su QT, Sun YH, Zhang XT, et al., 2022. A watermarking scheme for dual-color images based on URV decomposition and image correction. Int J Intell Syst, 37(10):7548-7570.

[25]van Essen D, Ugurbil K, Auerbach E, et al., 2012. The Human Connectome Project: a data acquisition perspective. NeuroImage, 62(4):2222-2231.

[26]Vaswani A, Shazeer N, Parmar N, et al., 2017. Attention is all you need. Proc 31st Int Conf on Neural Information Processing Systems, p.6000-6010.

[27]Wang CP, Wang XY, Xia ZQ, et al., 2019. Ternary radial harmonic Fourier moments based robust stereo image zero-watermarking algorithm. Inform Sci, 470:109-120.

[28]Wang N, Li Z, Cheng XY, et al., 2018. Diffusion weighted image reversible visible watermarking algorithm based on support vector regression. Proc 14th IEEE Int Conf on Signal Processing, p.1144-1148.

[29]Wen BY, Aydore S, 2019. ROMark: a robust watermarking system using adversarial training. https://arxiv.org/abs/1910.01221

[30]Wen Q, Sun TF, Wang SX, 2003. Concept and application of zero-watermark. Acta Electon Sin, 31(2):214-216 (in Chinese).

[31]Wu HP, Xiao B, Codella N, et al., 2021. CvT: introducing convolutions to vision Transformers. Proc IEEE/CVF Int Conf on Computer Vision, p.22-31.

[32]Xia ZQ, Wang XY, Li XX, et al., 2019. Efficient copyright protection for three CT images based on quaternion polar harmonic Fourier moments. Signal Process, 164:368-379.

[33]Yang FZ, Yang H, Fu JL, et al., 2020. Learning texture transformer network for image super-resolution. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.5790-5799.

[34]Zhan ZF, Cai JF, Guo D, et al., 2016. Fast multiclass dictionaries learning with geometrical directions in MRI reconstruction. IEEE Trans Biomed Eng, 63(9):1850-1861.

[35]Zhou B, Zhou SK, 2020. DuDoRNet: learning a dual-domain recurrent network for fast MRI reconstruction with deep T1 prior. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.4272-4281.

[36]Zhu JR, Kaplan R, Johnson J, et al., 2018. HiDDeN: hiding data with deep networks. Proc 15^th European Conf on Computer Vision, p.682-697.