Localized spatiotemporal modular ICA for face recognition

In this paper we have proposed a unique approach for face recognition based on modular Independent Component Analysis (ICA) with local facial features. The face images are segmented based on skin color using YCbCr color space. In this research work we have considered the samples of individual person which consist of sufficient number of images having pose variations, facial expressions and changes in illumination from Asian face database. The proposed method is based on local facial feature extraction after face segmentation. The local components such as eyes, nose, mouth (lips) are extracted automatically. These local components are used to obtain independent components. Using the independent components of these local facial components, the face recognition task is performed by ICA algorithms.