Covert exchange of face biometric data using steganography

In this paper, a high invisibility face biometric data transfer technique is proposed. The proposed method decomposes a face image into multiple frequency bands using wavelet transform. Each sub-band in the wavelet domain is divided into non-overlapping blocks. Then, local binary pattern histograms (LBPHs) are extracted from each block in each subband using only 4 neighbours to extract LBP code. Then, all of the LBPHs are concatenated into a single feature histogram to effectively represent the face image. Finally, the extracted face features are embedded in an image using one of the robust steganography techniques in order for them to be ready for transmission. PSNR between original and stego-image is calculated to measure invisibility of the system, while recognition rate of the system is calculated using Euclidean distance followed by a nearest neighbour classifier. The recognition is performed on the receiver side after extracting the embedded face features. The above strategy was tested on two publicly available face databases (Yale and ORL) using different scenarios and different combinations of sub-bands. Results obtained show that embedding LBPH features using our method will give higher invisibility whilst maintaining the recognition rate at the same level or better when compared with the original uniform LBP.