期刊名称：The World of Computer Science and Information Technology Journal
印刷版ISSN：2221-0741
出版年度：2015
卷号：5
期号：2
页码：6
语种：English
出版社：WCSIT Publishing
摘要：There are several research endeavors and results in aptly recognizing faces from video clips. However recognizing partially occluded and pose variant face images poses a greater challenge and concern for professionals in image and video processing fields. There are efforts underway for unearthing pragmatic algorithms and viable mechanisms for better recognition rate of occluded and pose-variant images. In this paper, a new method is proposed for efficiently recognizing human faces from videos with varying poses and partial occlusion. In the proposed model, a training dataset is created with faces of varying inclination and occluded region that are subjected to in-painting. A variety of features are being extracted using the discrete curvelet transform method from the captured faces in the training set and they are matched against the features from the videos for recognition. We have supplied the implementation results in this paper in order to demonstrate the efficiency of our approach and the algorithm leveraged. Keywords- video based face recognition; discrete curvelet transform; In-painting and face recognition.
