JPEG Compressed Domain Face Recognition: Different Stages and Different Features

JPEG compression standard is widely used for reducing the volume of images that are stored or transmitted via networks. In biometrics datasets, face images are usually stored in JPEG compressed format, and should be fully decompressed to be used in a face recognition system. Recently, in order to reduce the time and complexity of decompression step, face recognition in compressed domain is considered as an emerging topic in face recognition systems. In this paper, we have tested different feature spaces, including PCA and ICA in various stages of JPEG compressed domain. The goal of these tests was to determine the best stage in JPEG compressed domain and the best features to be used in face recognition process, regarding the trade-off between the decompression overhead reduction and recognition accuracy. The experiments were conducted on FERET and FEI face databases, and results have been compared in various stages of JPEG compressed domain. The results show the superiority of zigzag scanned stage compared to other stages and ICA feature space compared to other feature spaces, both in terms of recognition accuracy and computational complexity.