Research And Implementation Of The Relational Model Of Image Quality And Face Recognition Rate

With the rapid development of modern information technology, the demand for information security of the social from all walks of life is more and more strong. The realization of many application scenarios including online shopping and the network financial need to have a secure and efficient authentication technology support. And in numerous identification methods, face recognition with its own convenience, get more and more attention from people, and become the focus of researchers. At the same time, the rapid growth of the network speed, enables people to transfer photos, videos and other information on the Internet faster. Therefore, automatic image quality assessment also becomes more important with the popularity of Internet image transmission.As the input image quality will be greatly effects on the accuracy of face recognition, we can assume that there is a certain relationship between the image quality and the human face recognition accuracy. This paper makes it as the research object, and puts forward the relationship model of the image quality level and face recognition accuracy, and the face recognition accuracy can be obtained directly by the level of image quality. This paper first discusses the theoretical basis of the classic Viola- Jones face detection algorithm and the sparse coding classification model in the field of face recognition, as well as the QAC algorithm in the field of image quality assessment. And verify the effectiveness of these algorithms through experiments.Secondly, it studies the types of noise affecting the image quality, and make separate researches on the performance of image quality assessment algorithm of QAC with gaussian noise and gaussian blur. It can be concluded that the QAC algorithm can fit the human eye subjective visual perception well. Thirdly, it uses QAC ratings as a measure of the level of image quality scale, and studies the correspondence between the level of quality and recognition rate of both Gaussian noise and Gaussian fuzzy images, and eventually constructs the relational model on the face recognition rate and QAC score. Last but not least, in order to accurately divide QAC score range into several levels of image quality, this paper introduces the concept of fuzzy set theory, and proposed a "fuzzy interval". It uses the membership function to describe the interval boundaries, and more accurately reflects the quality of people’s subjective perception of the image, thereby further improving the relational model of the image quality and the recognition rate, makes it easier to accept the mapping between image quality and the face recognition rate. Experimental results show that recognition rate corresponding to the high-quality image is higher, and as the degradation of the image quality, recognition rate also decreased significantly.In this paper, a quantitative model of the relationship between image quality and recognition is achieved. It can be applied to guide the rational choice of imaging capture device.