一种用于图像认证的感知哈希方法

Perceptual Hashing Method Use for Image Authentication

随着大量多媒体应用的发展,数字图像增加了被非法操作和篡改的危险,这使得图像认证技术受到了越来越多的关注。感知图像哈希是图像内容或特征的简洁表示,它适合应用于图像内容认证和图像索引。提出了一种组合非负矩阵分解和奇异值分解的感知哈希方法,图像首先被分割成依赖于密钥的重叠图像块,每块进行非负矩阵分解得到表征图像局部特征的系数矩阵,然后对其进行奇异值分解取最大的奇异值矢量构成哈希序列,实验表明建议的方法对JPEG压缩、图像滤波等内容保持操作具有较好的稳健性,同时可实现定位恶意篡改的能力。

Digital images are easy to manipulated illegal and tampered with the development of a lot of multimedia applications. Image authentication techniques have recently gained great attention to the academia as well as the industry. Perceptual image hashing is the compact representation of image content or features. It is a technique suitable for use in image content authentication and image index. A novel perceptual hashing method based on non-negative matrix factorizations （NMF） and singular value decomposition （SVD） was proposed. The image is divided firstly into key-dependent overlapping blocks. Each block runs NMF to obtain the coefficient matrix that capture the local features of image. Then the coefficient matrix run SVD and use the biggest singular vectors form hash vector. Test results indicate that the proposed method is robust against content-preserving modifications such as JPEG compression and at the same time, is capable of localizing image malicious tampering.