Tamper Detection using Self-Generating Watermarks based on Local Binary Patterns

Image watermarking is a technique used for versatile applications. This paper proposes a blind spatial domain fragile watermarking technique that aims at detecting tamper and distortion caused due to intentional or unintentional attacks. The image is first divided 4 equal sections and 2 pairs are formed. These pairs are together used for tamper localization. Each section is then divided into $4 \times 4$ blocks. For every sub-block, the average of 4 inner pixels is calculated. Using this average value, local binary pattern (LBP) is generated for the sub-block using the 12 boundary pixels. A 4-bit watermark is self-generated from the LBP and embedded into the sub-block. The same watermark is also embedded into corresponding section pair to ensure better tamper detection. The method is tested against a variety of attacks like copy-paste, copy-move, blurring, cropping, etc. and shows efficiency in detecting tampered regions.