Study On The Method Of Eye-locating Based On SIFT Algorithm

Face recognition is the most popular research orientation in patterm recognition realm inlately decade. During these years, many new achivements in module and algorithm ofrecognition come out to the public.All of these shows that face recognition in its research parthas a boom futrue.In the processing of face recognition,given the human eye feature is themost important part of the facial features and it has ample visual features and good postion inthe face area, eye-locating make a huge contribution to both face recognition and facedetection.So not mention speed up the detection and recognition it also reduce the complexityof recognize algorithms ,all of the facts represents that eye-locating has a greatful meaning forprocessing the human face information. After elaborately research on two methods of locatingeyes in face which based on correlation and template matching, by using the SIFT featurewhich has particular advantage in describing the local features,this essay gives a new waywhich based on the features fetched by SIFT algorithm to locate eye region in the face image.The locating method of correlation mainly use the convolution to locate eyes in the face bycaculating the image information which has converted from time region to frequencyregion.Template matching always get a little template first then comparing the source imageand template image and come out the result.The method based on the SIFT feature whichusing its advantage in local feature description and combining with coordinate information tolocate eye area in the face image.The images which used to locate eye in the test are captured under the standard light sourcewith high resolution of human face different from the orl face database which used inexperiment.Firstly converting color image to gray one then storing in computer. Followed byde-noises treatment and histogram transformation,then it is suitable for various lightconditions and make the input image much better with higher contrast.After that,usecorrelation,template matching and SIFT algorithm to get the features of the eyes in the faceimages,with the finish of caculating the features,then get the location of the eyes and turn outthe conclution that by the features have advantages in scale , position and orientation whichbeen carry out by SIFT algorithm , the method based on the SIFT algorithm is better inprocessing non-front image than other methods.Finally ,use the artificial neural network classify the features and implement the facerecognition with principal component analysis.