Design And Implement Of Image Morphing System Based On Facial Feature Detection

With the development of the Artificial Intelligence technology, the demand of intelligent system have been raised rapidly, Face Detection, as essential theme of Intelligent Interface and important branch of Pattern Recognition, has been received much attention in recent years. During the announced production of this field, a kind method of samples learning, such as Adaboost, based on statistical models, gives out a good result.This thesis is based on subproject of Guangdong Science Center Project, collaborated by HUST and Guangdong Province, the number of subproject is "SS404-Magic Mirror-D21". From the prospective of practice, it describes the design and implementation of an image morphing system which based on detection of facial features automatically. Firstly, this system detects frontal face, using classifier of Haar feature and facial features automatically. Secondly, it takes advantage of field morphing algorithm and implements the image metamorphosis between the frontal face and animal's. Image morphing enables the fluid transformation of one digital image into another, it makes the median image has the features both of the source and destination images.The traditional morphing algorithms are mainly implemented on the whole image, it always demands that the matched information of morphing, such as points, segments or meshes must be assigned manually. The system described in this thesis has been improved greatly; firstly, design and implement an algorithm of image matting make the system much better in vision effect. Secondly, we figure out the contour of image accurately by using edge detection, design the map function of source and destination image and use polygon to simulate the real contour, which make the effect of morphing much more fluid, and the morphing algorithm run much faster. Finally, this system makes use of technology of face and facial feature detection; it detects the necessary information to morphing, and overcomes the shortcoming in traditional algorithms. As a result, the system can run automatically.This thesis depicts two significant topics of this system, face detection and image morphing; it discusses implementation of relevant technology, and has a great meaning in practice.