Publications

No-Reference image quality assessment for face images is of high interest since it can be required for biometric systems such as biometric passport applications to increase system performance. This can be achieved by controlling the quality of biometric sample images during enrollment. This paper proposes a novel no-reference image quality assessment method that extracts several image features and uses data mining techniques for detecting the pose variation problem in facial images. Using subsets from three public 2D face databases PUT, ENSIB, and AR, the experimental results recorded a promising accuracy of 97.06% when using the RandomForest Classifier, which outperforms other classifiers