Research On The Methods For Face Recognition Based On Sketch

Face recognition based on sketch, a key technique for individual identityauthentication, is more difficult than face recognition based on photo. This techniquewill find wide utilization in case-solving, public security and suspect-searching.Especially when the photo image of a suspect is not available, the best substitute isoften a sketch drawing based on the recollection of an eyewitness. Or some suspects'photos are timeworn and in poor quality and unrecognizable. In these cases, thistechnique becomes very important.This paper mainly studies the sketch based face recognition. The traditionalgeometrical and Eigenface(PCA) face recognition method are employed to recognizethe face sketch. We have analyzed the recognition result and improved these methods.First, the features of the photos and sketches are extracted by tow geometricalmeasures. The experimental results show that deformable template model achieves agood recognition accuracy of 77.5%. Second, a new gray-filled algorithm is introducedto significantly reduce the difference between photos and the corresponding sketchesby a transformation from a sketch to a photo, thus effective matches are obtainedbetween the two images. Finally, the PCA method is used for representation andrecognition of the gray-filled sketch face images. The experimental results demonstratethat the method can achieve an 82% recognition accuracy and the gray-filled algorithmis effective.