Based On The Average Energy And The Realization Of Lbp Face Image Quality Evaluation

With the rapid development of science and technology in information age, the computer and network technology have been being renewaled, which had expand the application field of digital image; modern people have an increasing demand for remote conference, e-commerce, distance education techniques, and related technologies have gained much attention and been developed continuously. Among all of these, face image recognition is a really active research direction in the field of pattern recognition research. And in safety, trade and economic fields, it has broad application prospects, such as criminal investigation, certificates verification, population control and video surveillance, etc. In order to meet the needs of the consumers, a safe, reliable, and efficient identification system had become urgent needs of related people, so as to provide them assured security, free their own rights and interests from unauthorized violation.However face image will be distorted in different levels and types during the process of acquisition, compression, processing, storage, transmission, and display process, and considering the particularity of face image, the person’s expression, posture of head, eyes open or closed, different kinds of decorations, hair on face and many other kinds of factors will affect the quality of face image. Low quality face image will bring negative impact to the performance of face recognition system inevitably, so establishing of an effective image quality assessment system is of great significance in the related fields of image processing and application.The quality of facial image has great impact on performance of face recognition system and other relevant systems, so people paid more and more attention on the research of face image quality assessment gradually. Due to the particularity of face image and numerous factors that have influence on the quality of face image, the evaluation methods of face image differ form the the ways of traditional image quality assessment greatly.In this paper, the take two main factors into consideration:the blurness and effects caused by illumination change. For Image blur, we define a concept of average energy of the image, use the value of the average energy distribution in frequency domain as the measure of the extent of blurness of face images. For the factor of illumination change, we use the symmetry degree of left-right of the face image as a reference to evaluate the face quality. At first, symmetrical subwindows of good quality image and poor quality images are selected to extract LBP(Local Binary Pattern) feature; these LBP features extracted are taken as samples for training and testing using AdaBoost algorithm, and some features that represent image symmetry most are selected after the process of AdaBoost; these selected features correspond to different sub-windows, we evaluate image quality scores by calculating the difference of these left-right symmetry sub-windows’gray level histogram.These methods have been proven effective through experiments.