Neural network method for detecting fuzzy images of faces

V. G. Spitsyn, Yu V. Savitsky, A. B. Kaziev, Yu A. Bolotova

Результат исследований: Материалы для типов конференцийДокументрецензирование

Аннотация

A method for estimating the degree of blurring of an image of a person based on the apparatus of convolutional neural networks is proposed. The possibility of using the original image, the modular component of the frequency spectrum and the phase componen t of the frequency spectrum of the original image as input to the neural network is investigated. The method was tested on an own image base of faces obtained from an IP camera in real conditions. The proposed method is compared with the known method of estimating the blur based on a quantitative analysis of the modular components of the frequency spectrum of the image. On the collected test sample, the proposed method showed the recognition accuracy of 98.57%, the method based on the quantitative analysis of the modular components of the frequency spectrum showed an accuracy of 77.12%.

Язык оригиналаАнглийский
Страницы238-241
Число страниц4
СостояниеОпубликовано - 2018
Событие28th International Conference on Computer Graphics and Vision, GraphiCon 2018 - Tomsk, Российская Федерация
Продолжительность: 24 сен 201827 сен 2018

Конференция

Конференция28th International Conference on Computer Graphics and Vision, GraphiCon 2018
СтранаРоссийская Федерация
ГородTomsk
Период24.9.1827.9.18

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Vision and Pattern Recognition

Fingerprint Подробные сведения о темах исследования «Neural network method for detecting fuzzy images of faces». Вместе они формируют уникальный семантический отпечаток (fingerprint).

Цитировать