Tampering localization and self-recovery using block labeling and adaptive significance
Abstract
This paper proposes a scheme for localization and restoration of image tampered regions using block labelling and adaptive significance. To generate the watermark information which includes authentication data and re covery data, we propose a block coordinate labelling method, which extracts the exact coordinate position in formation of each block, while the recovery data is composed of Block Adaptive Significances (BAS) and bitmaps, which are composed of high and low adaptive significance. To detect the tampered area more effectively, we propose a dual detection approach that combines the block-based labeling (BBL) and pixel-based labeling (PBL). We embed the authentication data into each pixel in the block sequentially and embed the position coordinate information of the block into the whole image in ascending order. The PBL approach can be used to rapidly complete tamper detection when the requirements for PBL are satisfied, whereas the BBL is used to increase the possibility of successfully detecting tampering if the conditions are not satisfied. Furthermore, we propose a block-level partially symmetric mapping and apply it to self-recovery bits in block units, thereby reducing the possibility of recovery bits being lost. The experimental results show that in our scheme, the average precision reaches 86.70%, which is 4% higher than the existing results, and the average F1score reaches 92.02%, which is 2% higher than the existing results.
Qiyuan Zhang
Phd student (since 2021.8)
Xiaochen Yuan
Associate Professor
