Very low resolution face recognition using fused visual and texture features

The recognition of very low-resolution face images are still challenging due to the absence of sufficient features. To solve this problem in this paper, a novel hybrid visual and texture feature extraction method is proposed. Here, the multiple size discrete cosine transform (mDCT) is used to extract the visual features and gray-level co-occurrence matrix (GLCM) is used to extract texture features. These visual features and texture features are fused to form the hybrid feature leading to improved the feature space. Moreover, feed forward neural network is incorporated along with selective Gaussian mixture models (sGMM) for matching the features. The public Georgia Tech face database was used to perform the experiment. The experimental results have shown that the proposed method achieve better performance than the existing methods.