Optimization of the Data Representation Integrated Form in the Viola-Jones Algorithm for a Person’s Face Search

Abstract

It is shown that the task of a person’s face recognition is one of the most popular at the present time. Its effective solution determines the reliability of many modern systems of personal identification, security, access control, and video surveillance at thermal and nuclear power plant stations. The speed of facial recognition algorithms is one of the key factors that determine the possibility of using for solving specific practical problems. It is shown that one of the methods for increasing the speed of facial recognition algorithms is the use of an integral form of data representation. The Viola-Jones algorithm is identified as one of the classic approaches to solving the problem of face recognition. The main limitations for using the classical integral form of data representation are shown. Rotation of the head is identified as the main factor leading to a significant increase in the error in determining the brightness indicators for the most informative areas of the face. An approach based on the use of modified integrated forms for image frame data representation is proposed. These forms are oriented to the fast processing of information in the presence of a significant rotation of the head. The most important range of possible changes in the angle of rotation of the head is identified. The data structure is considered in case of using the modified integral form. Experimentally confirmed the possibility of using a limited number of modified forms of data representation while maintaining the high reliability of the work of the human face recognition algorithm.