Local Non-Negative Matrix Factorization as a Visual Representation
ICDL '02 Proceedings of the 2nd International Conference on Development and Learning
Rotation and scale invariant texture features using discrete wavelet packet transform
Pattern Recognition Letters
Non-negative Matrix Factorization with Sparseness Constraints
The Journal of Machine Learning Research
Face recognition using fisher non-negative matrix factorization with sparseness constraints
ISNN'05 Proceedings of the Second international conference on Advances in neural networks - Volume Part II
Hi-index | 0.00 |
In this paper, a feature extraction method is proposed by combining Wavelet Transformation (WT) and Non-negative Matrix Factorization with Sparseness constraints (NMFs) together for normal face images and partially occluded ones. Firstly, we apply two-level wavelet transformation to the face images. Then, the low frequency sub-bands are decomposed according to NMFs to extract either the holistic representations or the parts-based ones by constraining the sparseness of the basis images. This method can not only overcome the the low speed and recognition rate problems of traditional methods such as PCA and ICA, but also control the sparseness of the decomposed matrices freely and discover stable, intuitionistic local characteristic more easily compared with classical non-negative matrix factorization algorithm (NMF) and local non-negative matrix decomposition algorithm (LNMF). The experiment result shows that this feature extraction method is easy and feasible with lower complexity. It is also insensitive to the expression and the partial occlusion, obtaining higher recognition rate. Moreover, the WT+NMFs algorithm is robust than traditional ones when the occlusion is serious.