A linear regression model for estimating facial image quality

The quality of biometrie data has a strong relationship with the performance of a face recognition system. The accuracy of automated face recognition systems is greatly affected by various quality factors, such as illumination, contrast, brightness, and blur. Therefore, an effective method is needed, which can characterize the quality of the facial image by fusing different quality measures into a single quality score. In this paper, we propose a novel quality estimation method based on linear regression analysis, to model the relationship between different quality factors and corresponding face recognition performance. A practical set of quality measures is used to estimate the quality scores. The linear regression model adjusts the weight of different quality factors according to their impact on recognition performance. The facial features are extracted using Local Binary Pattern (LBP) and k-nearest neighbor (KNN) classifier is used for the classification purpose. The prediction scores generated from the model is a strong indicator of the overall quality of a facial image. This model has many applications, for example, saving the processing time and improving the face recognition accuracy during enrollment processes by discarding poor quality images. The residual error of the regression model is 4.29%, and considering 0 and ±1 error between original response value and the prediction value results in a very high accuracy of 94.06%.