Quality enhancement of low-resolution face images

Image processing of human face is one of the major problems in machine vision science. Since direct outputs of surveillance cameras usually have low resolution and they are not appropriate for face recognition, it is important to improve the resolution of images extracted from surveillance cameras. This research focuses on resolution refinement of very low resolution (VLR) images. The main objective of this study is to refine the resolution of face images and to estimate high resolution image of human face in different poses which not even a low resolution image of that pose exists. For this purpose a method based on reconstruction or estimation of the desired image by using other human face images, stored in a database, is proposed. This method uses a scoring system based on the importance of different facial areas to achieve multi view face images.