Study On Face Images Automatic Cropping

Face image matching and automatic cropping is one of the hot research topic in the field of digital image processing. Compared with the traditional manual cropping, automatic cropping has greatly higher working efficiency, and it is indispensable in the internet and the application of artificial intelligence. As one of the key technology of face image automatic cropping method, face detection is increasingly urgent demanded in national and military security, financial, judicial, customs, civil aviation, the application of civil, social security and other fields.In this paper, the face region can be located according to the face detection technology. Based on the relative position of the center of the face in the sample size, and relative center location of the face in the image, the system can automatically match for the size and location of the cropping area of the picture to be cropped. And it can have the images automatically cropped at last. The main research work is as follows:(1) Choose the appropriate skin color segmentation YCCb color space after comparing several color spaces. Study two kinds of region segmentation algorithms based on color of skin, the ellipse skin color model and extensive skin model, and based on the rate of change of skin color pixels with the threshold, presents an improved adaptive algorithm, effectively solves the problems that traditional fixed threshold segmentation algorithm is lack of flexibility and ellipse skin model has poor robustness under uncontrolled lighting condition, providing a good basis for face detection.(2) Study the face detection of color images. In this paper, the mathematical morphology are studied, and cascade operation expansion and corrosion is used to eliminate noise and smoothing the contours of the image, obtains the very accurate skin color area. And then through five dimensions such as the hole in the area, regional external rectangle frame aspect ratio of the number, face region area, an area of screen face region, an face area of share and roundness to extract face region. Besides, it contains extensive studies of human eye and lip detection, and that based on these facial features, checking the face region to get a more accurate face region.(3) Study the position of the center of the face images in3different kinds such as a single, double and more people, calculates the relative center location of the face in the images by average method and the geometrical center point method. And based on the rules that in the template image and the ones to becropped, the proportion of the face in the area to be cropped are equal, and the area to be cropped of the template image and cropping images are fairly uniform in size and shape, this paper puts forward a method of human face image automatic cropping which has better cropping effect than conventional method that based on fixed position and size.(4) Abstract the core algorithm of the program into a dynamic libraries, presente friendly interface of dynamic libraries.