NEURAL NETWORK BASED FACIAL IMAGE QUALITY ESTIMATION SYSTEM AND TRAINING ANALYSIS

Face quality assessment has become a significant research area because of its importance in the task of face detection and recognition in unconstrained images from videos. In the surveillance video sequence, we get the imagery of people moving freely around the camera and, their face or head pose is generally not frontal and faces captured in videos may be of varying quality, it becomes important to discard the low quality faces and do the processing of high quality faces only. Seminal viola-jones method has been used to locate the facial landmarks. The error function computed as difference in location of measured landmarks and calculated landmarks has been used as the output of the neural network. The proposed method has been tested to give reliable quality estimation of the facial images. This work elaborates facial features extraction and quality estimation of extracted facial images. The details have been given regarding the facial features used for quality estimation of facial features, the output function used, and training of the neural network, which is used to generate the output face quality estimation score.