The Research On 3D Facial Reconstruction

Information that human being perceive is mainly gotten through vision.Three-dimensional reconstruction is always one of the hottest research direction in computer vision field.It is the technology to research how to get three-dimensional information of object from the two-dimensional information of object.This paper discusses two aspects of the three-dimensional reconstruction,one of them is 3D reconstruction based on medical CT slices and the simulation of plastic surgery,the other is 3D reconstruction from photographs of human face. The use of Computed Tomography(CT) and Magnetic Resonance(MR) equipment has brought the revolutionary transform of diagnosis and inspection of disease.These systems have offered doctors rich image information of organization and tomography of organ.Doctors can take diagnose more visually,quickly,accurately.How to carry out 3D reconstruction of these slices has more and more attractted the attention of researchers.This paper has studied the method of 3D reconstruction based on medical images and its related application.We get a sequence of images from medical equipment,reconstruct 3D entity and carry out operation with the 3D entity model,satisfying the requirement of clinical application. 3D medical image reconstruction denotes that reconstructing three-dimensional object from a series of two-dimensional medical images based on scientific calculation technology and the manipulation using computer.Combining clinical application on the foundation,we can realize operation imitation,operation programming and assistant diagnosis.At the same time,it promotes development and application of image processing technique and visualization technology. 3D medical image reconstruction includes these major contents:pretreatment of image input,image segmentation and surface reconstruction. Pretreatment In the process of acquisition of medical image data,noise is produced inevitably in each electronic equipment.The purpose of pretreatment is filtering and smoothing,restraining noise,strengthening image features,raising ratio of signal-to-noise. Image Segmentation Now the research on medical image segmentation is major based on MRI or CT images.Since the data of medical image is bulky and the quality of segmentation asking higher,all automatic segmentation method or all handwork method is greatly unsuitable.Therefore general method adopts semi-automatic interactive segmentation method.The images after segmentation may be directly used in further measure and assistant diagnosis.On the other hand,the images may be used to reconstruct accurate three-dimensional model. Surface Reconstruction 3D medical image reconstruction is build the three-dimensional geometry model of the image sequence input,after segmentation and feature extraction.The most common-use three-dimensional geometry model is surface model.Surface model is generally created with plane flat especially with triangle flat.Enclosing surface can form a polyhedron,also named polyhedron model.The early research is major concentrated on contour connection,that is to rebuild model from the contours.One representative work is the method of joining the surface of object with triangle flat.Along with the appearance of CT and MRI,reconstruction method based on volume data has been arisen. In recent years,human-computer interaction technology increasingly becomes a center of current research.Face detection has attracked more and more attention and become a hot point of pattern recognition and computer vision.Face detection is to determine the location,size and other state of faces in a given image or image sequence.The input of face detecting system is the image that possibly contained person faces.The export of the system is determining whether or not there is person face in image as well as the parameter of the number of the face,location and measure.Face detection is the important link of the application of face recognition,viewpoint trace andimage compression.From 1990's,face detection algorithm has gotten extensive research.Many approaches have been put forward,such as nerve network,machine study,shuttering match,Hough alternation and color analysis. Skin color offers important gist to face detection.Skin color is an important and obvious feature of face.It has relative stability and differs from the color of the most background objects.According to color image the use of skin color is an important content in face detection.Skin color does not rely on the details feature of face.It is not sensitive of enlarging,reduceing,rotating and other small warp.This paper presents a method of face detection based on the segmentation of skin color. To separate face area from non-face area,a reliable model is needed.The model should be suitable to different skin colors and different illumination conditions.RGB is improper for skin model,because in RGB space tricolor not only represent color,but also express brightness.Because of the change of surroundings illumination,brightness makes the face detection become more complex.Thus the skin segmentation will be not reliable.To use the clustering feature of skin color,brightness information and chrominance information in color expression are needed to be separated.The color model used frequently is mainly:YCrCb,HSV,YIQ,etc.In this paper YCrCb space is chosen to be the mapping space of distribution statistics.It has the advantage that the influence of brightness is less and its distribution is two-dimensional independent.It can restrict the distribution area of skin color.In YCrCb model Cr expresses red and Cb expresses blue in chrominance.YCrCb model may also be converted from RGB model by linear change.Therefore its application is extensive. Experiment shows that:though the skin colors of different person possibly differ greatly,the discrepancy in chrominance is smaller than it in brightness.That is to say,skin colors of different person are often close very much on chrominance.But the discrepancy is great in brightness. Three-dimensional face modeling is usually made up of three parts:3D data acquisition,general 3D face model and specific face modeling.3D data acquisition belongs to computer vision field.In recent years along with the development of computer vision technology,researchers have developed...