Using a cell phone camera
Using a cell phone camera
Many face recognition algorithms perform well on databases that had been collected with high-resolution cameras and in highly controlled situations. However, they may not retain good performance in real life situations where there is a lot of variation in illumination, scale, pose, etc. In applications such as face authentication using cameras in cell phone and PDAs, the cameras may introduce image distortions (e.g., because of fish-eye lens) and may be used in a wide range of illumination conditions, as well as variation in scale and pose. An important question is which of the face authentication algorithms will work well with face images produced by cell phone cameras? To address this issue, we collected a face database at Carnegie Mellon University using a China mobile phone camera. In this paper, we evaluate and compare the performance of correlation filters for face authentication with Individual PCA [1] and FisherFaces [2] under various lighting conditions. Correlation filters are attractive for a variety of reasons such as shift in-variance, ability to accommodate in-class image variability, ability to trade-off between discrimination and distortion tolerance, and the fact that they provide closed-form expressions [3-5].
The rest of the paper is organized as follows. Section 2 provides some background on correlation filters. Section 3 gives details on the database collection process using a mobile phone camera and the pre-processing done on these images. Section 4 provides an evaluation of correlation filters using this database along with a comparison with Individual China cell phone and FisherFaces. Finally, conclusions are provided in Section 5.