说明：  提出一种基于图像认证的半脆弱数字水印算法。该算法具有较好的自适应性，采用改进的低通滤波器，根据人眼的分辨能力设定 梯度阈值。实验结果证明，该算法对常规图像处理操作具有较强的鲁棒性，能够抵御JPEG顶级有损压缩，同时对于恶意篡改攻击(擦除、扭曲、挤压、光照等操作)表现出脆弱性，可检测出篡改的痕迹和类型。
(A semi-fragile watermark algorithm for image authentication is proposed, which has adaptability, and uses improved low-pass filter. The threshold values of grads are set according to resolving power of human eyes. Experimental results show that the proposed algorithm is robust to common data processing, which can resist JEPG top-level compression, as well the algorithm is fragile to malicious attack, such as abrasion, distortion, extrusion, lighting, and so on. Moreover, the tampered sort and area of watermarked image can be reflected)