The Age Estimation Of Facial Image

With the development of human-computer interaction (HCI)，the research onfacial imagehas been extensive carried out in areas such as image processing, patternrecognition and computer vision.Facial images recognition on expressions,gender,age, race, etcproblem become a research hotspot.As its potential value and relativelychallenging characteristic,facial image about age estimation has attracted more andmore researchers’attention.In this paper,we mainly studies on the human facial image feature extraction andprocessing method about age.In the feature extraction stage，we analyzed local binarymode (LBP) operator, and using it forage extractiontexture feature,at themeantime,we discussed its improved algorithm.we divided the images intoblocks,and we dealt each blockwith uniform LBP operator.After the feature extraction,in order to reduce the influence of factors other than age, and make using of theknown category information, the data were adjusted, so as to reduce class interval,increase class space, and enhance the accuracy in age face image recognition.In the feature dimension reduction stage, the commonly used algorithm--theprincipal component analysis (PCA) dimensionality reduction was analysed.As PCAis one of the linear dimension reduction methods, and it selectthe maincomponentsfrom the global view, it may lose the principal components with smalleigenvalue.What’s worse,the lost components may have vital information for the ageestimation.The improved algorithm Sp-PCA was choosed.the theory of Sp-PCA isthat PCA is performaed on each of the subpattern sets, and the sub-feature are thencombined into a global feature.The Sp-PCAretained the local important information,and due to the parallel computing on each blocks, it improved the calculation andrecognition performance.Our work and contributions could be expressed as the followings:(1) we present the the local binary pattern (LBP) operator forhuman facial imagefeature extraction, and discussed its extended model.The LBP operator is performedon the equally-sized non-overlapping portioned subpatterns.(2) Relevant component analysis (RCA) isadopted for age feature matrixadjustment after feature extraction, the using of the prior knowledge increases facialimage age recognition accuracy.Sp-PCA algorithm is put forward.In the dimension reduction stage,the imptoved algorithm Sp–PCA was employed, it not only improvethe calculation speed, but also retained the local effective information age.Experiment results and comparison with other published method show that ourproposed combined method LBP+RCA+Sp-PCA improved the age estimationaccuracyrate...