光学技术, 2018, 44 (4): 435, 网络出版: 2018-08-30
基于多元频域变换与几何校正的彩色图像水印算法
Image watermarking algorithm based on multivariate frequency domain transformation and geometric correction
图像水印 四元Fourier变换 SVM模型 最大中心区域 几何校正 四元小波变换 水印提取 image watermarking quaternion Fourier transform support vector machine model maximum central region geometric correction quaternion wavelet transform watermark extraction
摘要
提出了多元频域变换与几何校正机制的彩色图像水印算法。对重力模型进行改进, 实现水印信息的加密, 利用四元变换处理处理载体图像的最大中心区域, 获取其相位与幅度信息;设计水印嵌入机制, 将水印信息植入到四元变换的幅度中, 借助逆变换, 输出水印图像;引入四元小波变换处理水印图像的训练样本, 获取其四元小波变换系数训练支持向量机, 完成水印图像实现准确的几何校正;利用四元变换, 建立水印提取机制, 复原水印图像。测试数据表明: 与当前图像水印方案相比, 所提算法能够准确校正几何失真图像, 具备更高的鲁棒性与抗盲检测能力。
Abstract
The new watermarking algorithm based on multivariate frequency domain transformation and geometric correction was proposed. Firstly, the watermark information was encryption based on gravity model, and Multivariate Fourier transformation was used to process the maximum central region of cover image for obtaining the phase and amplitude information. Then watermark embedding mechanism was designed to embed the watermark information into the amplitude parts, and watermark image was output by Inverse Fourier transformation. The training sample of the watermark image was processed by introducing the QWT to DWT coefficients to for training the support vector machine so that completing watermark image geometric correction. The watermark extraction mechanism was established to restore the watermark image based on the correction of watermark image. The test results show that this algorithm can accurately correct the geometric distortion image, and it has higher robustness and anti blind detection ability compared with the current image watermarking scheme.
吴庆涛, 施进发, 曹再辉. 基于多元频域变换与几何校正的彩色图像水印算法[J]. 光学技术, 2018, 44(4): 435. WU Qingtao, SHI Jinfa, CAO Zaihui. Image watermarking algorithm based on multivariate frequency domain transformation and geometric correction[J]. Optical Technique, 2018, 44(4): 435.