| The automatic face recognition is a technology that using computer to analyzethe human face images and extracting e?ective features from the human face images,then to recognize them. Compared with other biometric methods, face recognition ismore direct, friendly, nature, and users can use it without any inconvenience.Based on the research of forerunners, along with the research in the algorithmof face detection and recognition, we have a deep research in image preprocessing,features extraction, and training the classify. Aiming at the characteristics of humanface, several e?ective algorithms have been designed in this thesis to solve the problemsin face detection and recognition.First of all, in the image preprocessing, this study basing on Gentle boosting learn-ing algorithm put forward by people before, adds a method for histogram equalizationof the images input in the algorithm training stage and feature computing stage, whichcan effectively improve the quality of the images, strengthen the useful information ofthe images and be more benefit to extract feature farther.Secondly, in the feature extraction, the study uses a method based on patch part-based feature extracting, which gather the features of the images random in type ofpatch and compute the distance to the center of the image. Compared with others, thismethod can also gather the features roundly of the images in the complex environment,without standing on the shape of human face figure and the abrupt changes of the graylevel.Thirdly, using PCA(Principal Component Analysis)for face recognition, this studydo the histogram equalization and edge detection to the images firstly, and then rec-ognize the result of face detection using the Gentle boosting algorithm. At last, wevalidate the feasibility of the integration of them.In the end, a system with a image database is done to test the algorithm introducedabove. The result of the experiment indicates that the algorithms are beneficial to thedesign. |
PDF Full Text Request