Face image segmentation by detecting attentive regions of artificial neural network

Several powerful face recognition systems has been proposed in previous research[1]. However, the recognition results of them are usually affected by the background images behind the target face images. Therefore, these systems need to get clean face images without the background images for the correct recognition. This paper proposes a face image segmentation system which detects and segments edges of the face image from pictures. This system consists of a six-layered neural network which recognizes face images. If the neural network recognizes a face in the input picture, the system segments features of attentive regions of the neural network. The segmented feature is the edge of the face image without any background images.