关键词: 数字水印, Haar小波变换, 支持向量回归机, 遗传算法

Abstract: To improve the quality of watermarked image, an improved robust watermark algorithm based on Support Vector Regression (SVR) and Genetic Algorithm (GA) was proposed. Following Haar wavelet transform, the wavelet coefficients which had strong similarity in image subband were adopted as feature vector, and then the SVR optimized by GA was used to build a wavelet coefficients direction tree model. The values of Mean Square Error (MSE) of the feature vector were compared to adaptively determine the information embedding position. According to the size between the prediction value and real value of the model, the watermark was embedded and extracted. The experimental results show that the proposed algorithm has strong robustness to common image attacks, even the Peak Signal to Noise Ratio (PSNR) can achieve 44.15dB with the embed capacity of 16384 bits. Thus, the proposed algorithm can resist watermarking attacks more effectively and it has high transparency under the situation with big capacity information embedded.

Key words: digital watermark, Haar wavelet transform, Support Vector Regression (SVR), Genetic Algorithm (GA