A Novel Image Signature Method for Content Authentication.

We proposed an image signature method for content authentication, which applies a hierarchical approach to construct an image signature. In the first level, DWT and DCT are used to extract image features; then these features are encrypted by using sub-keys that are generated by a cryptographically hash function. In the second level, Karhunen–Loeve transformation is used to reduce the signature length. The main features of the proposed method are as follows: (i) It achieves a trade-off between robustness and tampering sensitivity. (ii) It provides a tool for image tampering detection and tampering localization. (iii) It can be used to detect the thumbnail of the large image to improve detection efficiency. (iv) It provides the compact signature, and the signature length is independent of the image size. Experimental results show that proposed method is robust for content-preserving manipulations such as JPEG compression, adding noise, filtering and Gamma correction, etc. [ABSTRACT FROM PUBLISHER]