基于几何校正和LT码的鲁棒图像水印

摘要

针对目前多数鲁棒水印算法抗剪切、组合几何攻击性能较差,且嵌入容量小的状况,提出一种抗组合几何攻击的鲁棒图像水印算法.利用尺度不变特征变换(SIFT)算法从受攻击的图像提取特征点,与少量鲁棒性较强的原始特征点匹配,用RANSAC算法挑选匹配点对,根据特征点对间的坐标关系,用最小二乘法拟合仿射矩阵;根据仿射矩阵对图像进行几何校正;为提高算法的抗剪切性能,将水印信息先进行LT码编码,充分利用LT码的纠删性能抵抗剪切.实验结果表明,该算法有较大的嵌入容量,能够抵抗较多种类的几何攻击.%In view of the problem that many watermarking algorithms are poor under cropping and combined geometric attacks and show small embedding capacity,a robust watermarking algorithm was proposed.Feature points were extracted from the attacked image using the SIFT algorithm,and they were matched with a small amount of robust feature points extracted from the original image.The matching point pairs were further selected using the RANSAC algorithm.The least square algorithm was used to estimate the affine matrix of the geometric attacks according to the relationship between the matched feature points.The attacks were corrected based on the estimated affine matrix.To resist the cropping attacks,the watermark information was encoded with LT coding,to make full use of its excellent erasure correcting performance.Experimental results show that the proposed algorithm not only has high embedding capacity,but also is robust to many kinds of geometric attacks.