Face Selection for Digital Image Watermarking

Discriminative facial features vary from one image segment to another. A few researchers have studied most of the discrimination features for face recognition. According to them, face recognition concentrates on human face information without considering the specific subject, whereas biometric face recognition focuses on face-specific information by ignoring age and emotional information. However, previous works were concentrated on security aspect of face recognition by applying watermarking technology. As a consequence, the performance of face recognition technique was degraded due to side effect of watermarking on facial image. To solve this problem, a novel way is proposed to embed the watermark into less important parts of image by applying face selection technique. This paper used different state-of-the-art face selection to propose a new way to preserve most of the discriminative features of face image and to secure face image by applying image watermarking scheme. This proposed face selection scheme not only can enhance the performance of face recognition system, but also can save memory and computation time of the process of feature extraction. In this paper, face partition detection was done for each face image to extract person’s face. Also, watermark information is not embedded into the green channel of the RGB image due to human visual system (HVS). MOBIO Corpus was used for evaluation. The experimental outcomes showed an overall (21%) effectiveness of face selection scheme for face recognition and digital image watermarking. Hence, performance, memory and computation time can be improved by applying face selection for digital image watermarking.