This dissertation proposed that method to segmentalize of face image to separate face in background image and divided method to separate characteristic area in face image about edge detected of separated face image by get a differential image, and pro...
This dissertation proposed that method to segmentalize of face image to separate face in background image and divided method to separate characteristic area in face image about edge detected of separated face image by get a differential image, and proposed method to extract facial characteristic in wavelet conversion area and method to recognize face image using neural network.
Also, because face size of a person of fixed distance is resemblant almost from camera and detected characteristic area of four corners shape, a basis of knowledge for human face and consider of edge distribution face image.
And, run wavelet conversion and drew characteristic.
Run 2-levels wavelet conversion and 4-levels wavelet conversion about segmentalized facial characteristic area and detected each 92, 108 characteristic vectors with investigate coefficient distribution.
When detected characteristic using 2-level wavelet conversion, average RMSE for experiment image 0.0025 and using 4-level wavelet conversions average RMSE 0.0043 about same person that can yield exquisite characteristic vector more if use wavelet conversion.
Result that recognize after study to neural network, in case run cognition because using wavelet conversion, see the awareness rate of 100% about studying image and 92.18% about experiment image and expressed the awareness rate of 96% on the whole.