Algorithm to obtain anthropometric parameters in frontal images of human faces
Abstract
This article describes an algorithm developed under image processing techniques to automatically obtain relevant anthropometric parameters from frontal human faces. The algorithm first applies a Gaussian model of the skin to detect the individual face in the scene. Later, a set of morphological equations is applied to isolate or segment the interest regions of the face (eyes, nose and lips). Once these regions are detected, the algorithm calculates the reference point’s coordinates and finally they are used to obtain the anthropometric measures of interest. The implemented algorithm is capable to automatically detect 14 of the 18 referent points defined by Farkas for frontal faces with a mean square error less than 3 mm. With this set of points 23 parameters are measured in an average computation time of 13 seconds.
