Informatics

Combined method of face detection on images using Gaussian mixture model and Haar's cascades

D. M. Timoshenko

St. Petersburg State University, 199034 St. Petersburg, Russian Federation

Abstract: The well known Gaussian mixture model and the Viola–Jones algorithm are combined into one method which provides a high detection rate and a low false negative rate. A two-dimensional windowed discrete cosine transform is shown. Discrete cosine transform coefficients were used as learning features for a mixture model. Optimal parameters of Gaussian mixture model and the number of the first cosine coefficients are found out. The proposed method is compared with Viola–Jones approach on a large image database from social networks. Bibliogr. 17. Il. 4. Table 2.

Keywords: face detection, image processing, pattern recognition, Viola–Jones, Gaussian mixture model, cosine transform.

UDC: 519.6

Received: May 30, 2013