3D Face Reconstruction Based On Generic Model And Single Image

Three dimensional modeling is always one of the basic problem in Computer Graphics and Computer Vision. Being the important carrier of communication and emotion expression, human faces has been a favorite object to numerous outstanding modeling algorithm, for its specific generality and usability.Efficient modeling algorithm plays a critical role in enhancing the ability of PC to understand facial expression and enrich human-computer interacion at the same time, it also the meaningful reference to the other algorithms aim to different objects.However, it is a challengeable subject to build a vivid3D facial model, for the extremely complicated geometry shape, surface material, subtle changing of expressions and various profiles of human faces. Even if it is difficult to modeling, a mass of explorers still hold an active attitude in this area, and lots of great algorithms have been widely used in the field of commercial film、computer animation and games, live communication, medical diagnosis and facial recognition.Most of the traditional3D facial modeling based on extracting characteristics information from pounds of facial graphics to and then optimization the algorithm to finish regular models’modification. The method based on morphable model is a3D facial modeling algorithm appears in recent years, it has the low-complicated mathematic operation and steady output. Besides the characteristics talked before, it also can finish the modeling based on single graphic point to the characteristics and the problems of morphable model, this article studys modeling based on the single graphic, the main tasks include:(1) Based on active shape models’facial characteristics extraction.Biometric feature points positioning is always the basic problem in pattern recognition, and also the premise of model-rebuilding. This article emphasizes the way based on Active Shape Model (ASM) to get the plane feature point positioning.Through artificial marks on center of eyes and mouth to finish the original positioning of ASM global shape model. Although imported artificial intervene, this method reduced the complication of the algorithm and program.At the end, injected the modifying part points to the location of feature points, to ensure the accuracy of positioning.(2) Make the result of modified Hausdorff distance survey as the evidence of deformation of deformation.Generally speaking, Modified Hausdorff distance(MHD) used to survey the similarity of geometric shape between different point-sets, and the smaller of MHD value is, the more geometric similar. As human face, different facial organs have different interference values to facial similarity, and the eyes and facial feature provide more in similarity. Using weighing facial similarity to calculate MHD, then reverse the result by single-valued calculation, made a smaller MHD value to corresponding a larger one. Finally,using morphable model to calculate the3D feature points corresponding to the certain face to finish the changing of feature points from plane to3D.(3) Bringing up new method of vein mapping.This article divides the model-rebuilding into shape rebuild and vein mapping. The rebuild of Shape model is based on a plastic reshape general model. Then projection the rebuilded shape model, the result is a flat grid model which is same as the front of facial shape, then the zoom out and rotate flat grid model can be matching and corresponding to the facial part of the original graphic, called vein mapping. Finally, get the color value of every model apex of the original picture. The experiment result declares that the based shape modeling to finish vein mapping is feasible.(4) The application based facial-rebuilding model.This article’s facial-model reconstruction is emphasize the approximate of facial rebuild, not related to complex physiology dissection data, as the result, most of the applications is suitable to the environments which is not require high accuracy of model. Reconstruction model is given in the paper in surgical medical, facial expression simulation and general network in the application example.Through the deeply study in the areas talked above, this article realizes the3D facial modeling based on single plane graphic, and solves many critical problems in building models, obtains much scientific achievements.