Abstract
In the last decade face recognition has made significant advances, but it can still be improved by applying various techniques. The areas that have high promise of improvement are those that utilize preprocessing techniques. The main objective of this study is to improve the auto face recognition system performance using off-the-shelf image library. Face detection technique plays a significant role in recognition process. The process chains used to detect human face are those that comprises of color segmentation, localization using Haar-like cascade algorithm and geometry normalization. Subsequently, one half portion of the facial image was selected to be used as the calculated average half-face image. The high-dimensionality of the image value is further reduced by generating Eeigenfaces. This is followed by the classification process that was achieved by calculating the Eigen distances values and comparing values of image in the database with the captured one. Finally, the verification tests are carried out on images obtained from VidTIMIT database to evaluate the recognition performance of the proposed framework. The resultant tests from the data set yielded the following results: true acceptance rate at 91.30% and false acceptance rate at 33.33%. The obtained experimental results illustrates the proposed image preprocessing framework improves the recognition accuracy as compared to not applying it.