Aging Simulation Based On The A Sparse Constrained Non-negative Matrix Factorization Face Image And Face Recognition

Aging simulation based on facial image is a very important research topic in the fields of computer vision and image processing, as well as an important research aspect and technical difficulties in the face recognition field. NMFsc (Non-negative Matrix Factorization with sparseness constraints) algorithm not only maintains the additive trait of the original algorithm, but also makes data more local representative through adding sparseness constraints. Considering the local trait of the effects of age variation upon facial appearance, we explore and study the technique of aging simulation for facial images based on NMFsc in this paper, which will be finally applied in the field of face recognition across age progression. The major contents and relative achievements are listed as follow:1. A method of facial aging simulation based on NMFsc has been proposed.Facial images can be decomposed into basis matrices and coefficients matrices by NMFsc algorithm. The adjustment of sparseness for basis or coefficients matrices contributes to the establishment of facial feature model by means of subspace method, which has a strong ability of local representation for facial images. Afterwards, we use an improved prototype method to carry on the facial aging simulation. 2. We have come up with a facial aging simulation method based on residual image and wavelet transform.The residual image between NMFsc reconstructed image and original one has preserved high frequency information of the original image, which enables us to synthesize prototype face with more representative of age. Wavelet transform is capable of extracting high frequency sub images carrying characteristic of aging skin texture from given template image. It is useful to simulate facial images with much aging characteristic.3. A method of face recognition and face retrieval with age variance has been put forward.Apart from lighting, gesture and expression, variations in shape and texture of human faces due to aging factor would also affect the performance of face recognition system extremely. Therefore, we apply the aging simulation method proposed above into age-invariant face recognition and retrieval in order to simulate virtual samples with specific age to finally improve the recognition and retrieval ratio.