Studying the Performance of Different Transforms in Face Image Authentication Technology

The rapid spread of face image manipulation applications and the malevolent uses of the manipulated face images have been considered as an alarm for information security specialists. Inspired by the urgent need of face image manipulation reveal, various researches have been introduced to serve this purpose. The recent direction of developing face image authentication (FIA) technology has been directed towards combinations of various image processing techniques such as face area detection, image watermarking, tamper detection, tamper localization, and image recovery. In this paper, FIA scheme has been implemented based on various transforms such as Discrete Cosine Transform (DCT), Integer Wavelet Transform (IWT), and Slantlet Transform (SLT). The aims of this paper is to study the performance of these transforms in FIA technology and choose the one with the best results in addition to improving the embedding capacity. The first part of experimental work has been dedicated for testing the performance of the transforms which proved that the SLT obtained the best visual quality results. The second part of the experimental work have been dedicated for testing the SLT-based FIA scheme. The results proved that the scheme can successfully detect manipulated blocks in the face region and recover the original face region with good visual quality. The proposed FIA scheme proved its effectiveness compared to the state-of-the-art watermarking-based schemes; therefore, it is recommended for practical applications.