Murygin K. Face recognition based on a correlation methods

The object of research is a digital frontal face images that have vertical axis of symmetry. The purpose of research is solution of the problem of computer person identification based on frontal face image analysis. The research methods and equipment are theoretical and experimental analysis, personal computer. Theoretical and practical results consist in: there are proposed new and are improved exist methods of automatic face detection and recognition, is proposed method of solving of face detection and recognition tasks in single complex, the methods was developed can be used for applied tasks of access control and database searching by face images. Novelty of introducing are: proposed new approach for solving face detection and face recognition tasks in single complex that based on a determination of necessary discrete in scales used for face detection on the basis of required limitation on recognition error percent; the correlation method for the face detection was improved in direction of reduction of influence of between-class scatter of face images on the detection results, was developed two modifications of this method contained in comparison by parts and conversion to principal components that allows to increase the correct classification percent of the face and background images; the identification method based on the Gabor kernels were improved in direction of decreasing of the total recognition error by using of reduce set of scales; are proposed the identification method based on the union of Gabor kernels approach and the LDA methods that allows to get a new feature set that have smaller size and reduce the total recognition error. Field of using is development of computer systems for video observation, access control; face images database analysis; and the education processes.