3D Face Modeling And Applications Based On One Image

Along with the development of multimedia technologies, the3D face modelingtechnology becomes research hot spot in computer vision and computer graphics in thelast few years. The traditional3D face modeling methods need the information of severalface images, and then modify a general face model. Modeling3D face from one image is anew way which was proposed several years ago.3D face modeling method based onmorphable model is one of the best modeling at present. This paper concentrated on the3D face modeling from one font face image.Firstly, the face feature points was extracted.3D face database was builded from theface of CVL Face Database. Against the morphable model shortcomings of manyoptimization parameters, the complex objective function, poor matching efficiency andlow reconstruction accuracy, this paper used the improved principal component regressionalgorithm and spare morphable model for face shape reconstruction.Secondly, though opening up the face texture panorama face texture database wasbuild by mesh subdivision. The texture parameter was optimized by the improved particleswarm optimization algorithm to reconstruct the face texture. The3D face reconstructionwas implemented ultimately.Finally, this paper studied the application of3D model.3D model was applied to theface pose estimation. Linear regression algorithm was used to estimate the pose and thenthe3D face pose was revised to acquire the accurate pose value.This paper carried out simulation tests using Matlab for face modeling algorithm,improved some key techniques in face modeling, and implemented the3D facereconstruction from one image. At last we analyzed the existing problems, and providedresearch directions and some experiences for the further study.