A Reversible Image Authentication Method without Memorization of Hiding Parameters

This paper proposes a novel reversible image authentication method that does not memorize the parameters for extracting embedded authentication data from an image. The proposed method once distorts an image to hide data for authentication into the image, it recovers the original image from the distorted image unless tamper is applied to the image, i.e., reversible. By comparing extracted data and data generated from the restored image, this method detects image tampering and further localizes tampered regions by the unit of block. The proposed method extracts hidden data without memorization of parameters used in its algorithm. This feature makes the proposed method practical. Whereas any method memorizing parameters faces severe problems with storage and management of parameters, according to the increase in the number of memorized parameters that is caused by serving accurate tamper localization and/or by applying itself to a huge number of image collection, e.g., video sequences. Simulation results show the effectiveness of the proposed method.