Human Face Recognition Research Based On Gabor Wavelet

The human face recognition technology is lets the machine have human's intelligence, may remember, identify human's one kind of front technology. It unified the computer graphics, the computer imagery processing and the pattern recognition and so on many kinds of discipline domain. Along with the society to fast, the accurate automatic identification authentication actual demand is day by day urgent, the human face recognition already became the pattern recognition and the computer vision domain research hot spot. In has carried on the classification and in the summary foundation to the recent years some main person face recognition algorithm, this article proposed one kind new based on Gabor wavelet person face recognition methods, and to this method theory basis, the realization process and so on has carried on the comprehensive analysis. This method characteristic and the innovation place of summary is as follows:①The eye localization was the key link which the human face distinguished, this article proposed one kind of improvement symmetry transformation algorithm, might enhance the rate of accuracy which the human face image the eye located.②As carried on the comprehensive experiment to each direction and the frequency Gabor person face characteristic, draws the conclusion: Is all essential regarding the multi-posture person face recognition each direction characteristic information, may select the partial frequencies the Gabor filter to attribute the human face.③Selects the partial frequencies the Gabor filter to attribute the human face, uses the wavelet decomposition to the human face Gabor transformation image to carry on falls Uygur, to the human face different expression, the different direction all has the high recognition rate.④Uses the wavelet decomposition to the human face Gabor transformation image to carry on falls Uygur, is more effective than the traditional sampling law.Selects certain Gabor filter to attribute the human face, each person's human face image all has the different expression, in does not make the pretreatment, adjusts the human face criterion under the premise, the recognition rate may approach 100%.