In the traditional ICA method the entire face image is considered for holistic approach of face recognition, hence large variation in pose or illumination will affect the recognition rate profoundly. In this approach dividing the face image in sub-images, independent components are obtained on these sub-images and used for face recognition. Here we have explored modular ICA approach with partition of facial images as well as with local facial components such as eyes, nose and mouth. The face recognition task affects due to presence of noise in facial images. We have experimented ICA algorithms for reduction of noise from facial images so as to reduce noise effect. The research work presented in this book and methods proposed for face recognition are unique and definitely will provide new way of analyzing facial features. This will be a good contribution for research in biometrics and image processing field.