Eyes And Mouths Detection And Localization In The Images

In this paper, find the location of the face regions, then detect and locate the eyes and mouth in the existed face regions and target the important characteristics of the relevant points, such as: regional center, eyes corners, and mouth corners. Main research methods are as follows:(1)Propose a new approach to segment the face regions. Currently, the face regional segmentation mainly relies on skin color space in certain specific characteristics of the cluster. But the light of the face is easily impacted by the light of the environment, which causes uneven brightness of the face. And when there are backgrounds with similar color with the skin or existences of multiple skin regions, it is difficult to accurately positioning face region. This paper proposes using face contour and facial features show the distribution of triangular position to locate the face regions, which can well avoid the above-mentioned impact.(2)Propose a new approach to locate the eyes and their characteristics. In the binary images, to detect eyes in the first half of the face region, this can short the testing time. Using the Minimum Enclosing Rectangle (MER) to describe the contours of the regions, and locating the eyes regions based on the length and width ratio and the location, consistency and other constraints distinction between the eyes and eyebrows, hair region. And find the center of the eyes using the area centroid. Then, locate the exact position of the pupils corners using Harris corner detection algorithm based on the filling algorithm and morphological close operation.(3)Propose a new approach to locate the mouth and its characteristics. Changing the second half of the face regions into binary ones, which can much completely save the mouth contours. Then using the Minimum Enclosing Rectangle to describe the contours of the regions, and locating the mouth regions based on the length and width ratio and the location information, and find the center of the eyes using the area centroid. Finally, locate the mouth corners using Harris corner detection algorithm based on the filling algorithm and morphological close operation.According to many tests, the algorithm can easily and accurately detect the targets. Especially, the method of face segmentation well resolves the problems of that rely on the clusters of the skin. Using the filling algorithm and morphological close operation to help the harris corner detection algorithm detect the corners of mouth and eyes improved the accuracy. To detect the eyes and mouth respectively in the first and second half of the face regions can short the testing time and improve the accuracy.