数画像からの3次元ヘアモデルの生成

抄録

This research proposes image-based hair modeling algorithm that is effective at producing a realistic hair model for virtual characters in the world of virtual reality, movies, television commercials, and simulation games. To create a realistic and physically accurate hair model, the importance of hair cutting and bending must he considered. The problems of cutting and bending are solved through the use of multiple images of a human model. Multiple hair images taken from different directions provide the cutting criteria. To calculate the bending angle, hair volume is produced from three images of the head taken from the right, back and top. This hair volume represents the real space volume where hair is found and guides the bending process until the hair strands are inside the hair volume. The use of the hair volume together with the hair-area-outlines produces a hair model that is not only realistic but also closely resembles the human model in the picture.One main advantage of this hair model is the ability to model hair for specific individuals which is very difficult to achieve with other methods.