A neural network based human face recognition of low resolution images

In this work, a human face recognition algorithm based on Block-based Discrete Cosine Transform (BBDCT) and Extreme Learning Machine (ELM) is proposed for low resolution input images. We also investigate the effect of image resolution on the recognition rate of the proposed face recognition system. Furthermore to improve the low resolution input images, three interpolation schemes, namely, Nearest-Neighbor, Bilinear, and Bicubic, are used as a pre-processing step to obtain better recognition rate. The experiments are conducted on the ORL database to demonstrate the performance of the proposed algorithm.