Extracting Characteristic Face Points From 2D Image

Abstract

This work reviews the methods of finding head in the image, detection of characteristic face points and describes head and faceanthropology proportion. Given method use combinative geometry-based and image-based methods. This method extracts characteristicface points from 2D image and finds accurate coordinates. In the first step we extract head contour from background and find the areawith eyes. Image-based methods finds accurate pupils centers with iris. We get accurate lips and nose coordinates using Canny gradientmethod. The described method was tested by experiment. Accuracy of the method is 81 %. It is not very high accuracy, but it is enoughfor automatic head tracking and facial expression. Ill. 9, bibl. 7 (in Lithuanian; summaries in Lithuanian, English and Russian).