Image-based High Accuracy Face3D Reconstruction Algorithm Research

High accuracy face reconstruction has large and widely application in computergraphics, movies, games, medicine and teleconference etc. It is becoming a newfocous of computer vision. The highest accuracy and widely used3D technique isactive reconstructive method. However, because of the expensive of the equipment,the technique will not be promoted in more fields. A new passive method witch isalways called image-based reconstructive method becomes more and more popularfor its convenience and low requirement. This thesis studys an image-based highaccuracy face3D reconstruction algorithm.A new hierarchical dense matching method using Daisy is proposed in thethesis. The Daisy operator has high precision and high speed advantages, but thehuge memory consumption limited its application in the high resolution imagesmatching algorithm directly. A stratify of Daisy matching algorithm in highresolution images corresponding is proposed. An initial dispatity map has beenachieved with a hierarchical matching. And then a modified disparity refinement hasbeen done to get a high accuracy pionts clouds. At last, a high accuracy model willbe reconstructed useing Poisson Surface Reconstruction (PSR) method.The only inputs of our method is a paire of face calibrated images taking instandard illumination, and then a hich accuracy face3D model will be reconstructeddirectly. C++programing language has been used to realize our matching algorithm.Some images which are took in with digital cameras are needed to be inputs, andthen the reconstructive system will output a high accuracy model. Relative to otherface reconstruction methods, the results shows our algorithm is simple to implementand high accuracy.