Towards nonuniform illumination face enhancement via adaptive contrast stretching

A face enhancement has the potential to play an important part in providing satisfactory and vast information to the face recognition performance. Therefore, a new approach for nonuniform illumination face enhancements (NIFE) was proposed by designing an adaptive contrast-stretching (ACS) filter. In a more objective manner of achieving this, an investigation usage of CS function with adjustable factors value to summarise its influence on the NIFE is examined firstly. Secondly, describe a new strategy to cater for CS adaptive factors prediction using training and testing phases. A dispersion versus location (DL) descriptor was examined in the training phase to generate the faces feature vectors. Subsequently, a frame differencing module (FDM) was developed for faces label generations. In the testing phase, the approach was examined to recognise the DL descriptor and predict face label based vocabulary tree model (VTM). Thirdly, the VTM performance was examined by referring to the area under curve (AUC) score from the receiver operating characteristic (ROC). The face quality measurement was evaluated via blind reference based statistical measures (BR-SM), blind reference based DL-descriptors (BR-DL) and visual interpretation of the resulting images. The BR-SM performed through calculating the EME (Measure of Enhancement), EEME (Measure of Enhancement by Entropy), SDME (Second Derivative like Measure of Enhancement), SHP (Coefficient of Sharpness) and CPP (Contrast per Pixel). In addition, by using DL scatter, the BR-DL handles the specific relationship with regards to the local contrast to local brightness within the resulting face images. Four face image databases, namely Extended Yale B, Mobio, Feret and CMU-PIE were used. The final results attained prove that compared to the state-of-the-art methods, the proposed ACS filter implementation is the most excellent choice in terms of contrast and nonuniform illumination adjustment as well as providing images of satisfactory quality. In short, the benefits attained proves that ACS is driven with a profitable enhancement rate in providing tremendous detail concerning face recognition systems.