Automatic extraction of face contours

Abstract

Accuracy of the human face contour can help evaluate the facial orientation to retrieve facial features to identifying or verifying a human facial expression. Automatic extraction human face contour algorithm with high accuracy is necessary. In this paper, a novel flowchart of the face contour extraction algorithm was proposed for improving accuracy of face contours. Poisson Gradient Vector Flow (PGVF) Active Contour Model with an edge map is used to extract face contours. The edge map of a whole face contour is constructed by the Divided-and-Conquer technique and Canny edge detector. Genetic algorithm is implemented to automatically find the parameters of Canny edge detector. Three datasets with images and temporal sequence images have been tested for evaluation of the proposed algorithm. The expermental results demonstrated the algorithm can obtain accurate face contours