SVD based fragile watermarking scheme for tamper localization and self-recovery

This paper proposes a novel fragile watermarking approach for digital image tamper localization (TL) along with the self-recovery (SR) capability. The host image is first divided into blocks of size 4 x 4 and then, singular value decomposition is performed on each block. The trace of singular matrix is used to compute the TL bits for each block. These 4 x 4 blocks are further divided into blocks of size 2 x 2 to compute the SR information. This SR information is also used for 2nd level of tamper detection in order to enhance the resolution of TL. Proposed scheme implements a self-embedding of TL bits and mapped embedding of SR bits and for that, block mapping is performed. The encrypted combination of both the codes (TL of parent block + SR of mapped block) is inserted in the 1st and 2nd least significant bits of the parent block. Encryption of watermark bits (unique for each watermarked image) helps it to sustain the block based attacks (vector quantization, collage and four scanning attack). Whereas; two level of authentication helps it to sustain content only and constant average attack. The scheme provides an average detection rate of more than 99.5 % and average recovery of 28 dB for a 50 % tampering. The proposed scheme shows a superior performance, when it is compared with the earlier proposed schemes.