Differential Abnormality-Based Tampering Detection in Digital Document Images

Copy-move and splicing are familiar operations for tempering digital document images. In this paper, we propose a simple yet effective method for detecting such document image tampering blindly. The spatial-domain differential abnormality is often discovered along image splicing boundaries under the assumption of being without or with weak post-processing. In order to capture such a forensic clue, the row- and column-wise first-order image differences are computed and then binarized via thresholding. Such binarized difference maps are median-filtered and fused to resist background noise. The quadrangle structure consisting of four latent boundary lines is searched particularly via adaptive sliding windows within the fused map. Preliminary test results on an elaborately collected image dataset verify the effectiveness and efficiency of our proposed document image tampering detection method.