Шапошников Д.Г., Головань А.В., Гизатдинова Ю.Ф., Литовченко Е.В., Гусакова В.И.
НИИ Нейрокибернетики им. А.Б.Когана РГУ
The algorithms and procedures for classification of the most informative regions (MIRs) of facial images have been presented. Besides, method for estimation of facial racours based on determination of mutual geometric relations between the MIRs has been described. Computer simulation shown high effectiveness of the proposed methods while processing images from ORL database and detection of the MIRs by operator (p=0.97 for classification and 0.95 for racours estimation). While automatic detection of the MIRs by the cascade method the effectiveness was lower (p=0.7 for identification of eye regions and 0.87 for racours estimation). Possible way to improve the performance of elaborated method is discussed.