Image tamper detection and multi-scale self-recovery using reference embedding with multi-rate data protection

Abstract

This paper proposes a multi-scale self-recovery (MSSR) approach to protect images against content forgery. The main idea is to provide more resistance against image tampering while enabling the recovery process in a multi-scale quality manner. In the proposed approach, the reference data composed of several parts and each part is protected by a channel coding rate according to its importance. The first part, which is used to reconstruct a rough approximation of the original image, is highly protected in order to resist against higher tampering rates. Other parts are protected with lower rates according to their importance leading to lower tolerable tampering rate (TTR), but the higher quality of the recovered images. The proposed MSSR approach is an efficient solution for the main disadvantage of the current methods, which either recover a tampered image in low tampering rates or fails when tampering rate is above the TTR value. The simulation results on 10000 test images represent the efficiency of the multi-scale self-recovery feature of the proposed approach in comparison with the existing methods.