结合保角变换和Zernike矩的稳健图像Hash

Perpetual Image Hashing Based on Conformal Mapping and Zernike Moments

提出一种用于图像认证和篡改检测的稳健图像Hash方法.根据密钥对图像进行伪随机分块,将图像块规格化为大小相等的正方形并进行保角变换,使各图像块映射为单位圆.对各圆形图像区域计算修正的Zernike矩,连接所有Zernike矩的幅度和相位,基于密钥置乱得到所要求的图像Hash.与直接对正方形内切圆计算Zernike矩的方法相比,保角变换使图像块四角纳入Hash所表征的图像内容,因而有效改进了篡改检测性能.图像分块使Hash具备篡改定位的能力.实验表明该方法具有很低的碰撞概率和良好的感知稳健性,与其他方法相比,碰撞率和检错率较低.

We propose a perpetual image hashing method applicable to image authentication and tamper detection.The input image is pseudo-randomly partitioned into blocks,which are resized to become squares of a standard size and mapped to a unit circle by conformal mapping.Zernike moments of the circular-shaped ＂images＂ are calculated,and then modified by multiplying an exponent to achieve phase rotation-invariance. All amplitudes and phases of the modified Zernike moments are concatenated to form an intermediate Hash sequence.The final Hash is produced by pseudo-randomly permuting the intermediate Hash.The purpose of conformal mapping is to include information contained in the corners of square image blocks into the extracted image Hash so that the tamper detection capability is significantly enhanced.Experimental results show that the proposed method is robust against most content-preserving image processing operations.Image blocking allows the tampered regions in the image to be correctly located.Compared to some other techniques,the method has low probabilities of collision and less error in tamper detection.