3D Garment Modeling Based On Multiple Fusion Of Templates And Characteristics

3D garment is widely used in many fields of virtual characters, virtual fitting, online games, film and television animation, etc., and becomes a hotspot in computer graphics and computer vision. The method of image-based virtual clothing modeling attracts a lot of attention, because of its convenience of obtaining information resource and the realistic effect of modeling. So the research of image based virtual clothing modeling method has the vital significance for promoting the application of virtual clothing and the development of related industry.Based on analyzing the existing garment modeling technology, template technology and fusion technology, this paper proposes a method of 3D garment modeling based on the multiple fusions of characteristcs and templates. We use image segmentation technology to realize garment region segmentation, mix the image shape contour features and style template together by partition, and realize the information extraction of garment parts; we fuse the clothing style features, the loose quantity characteristics and size characteristics with the human template step by step, and obtain the template of garment modeling based on the human body template; we fuse image shape contour feature, clothing characteristics of size and color texture features with the modeling template hierarchically, and then realize the realistic modeling of 3D garment.We present an extract method of garment parts based on partition fusion between the structure characteristics and the clothing style template. Attaining the structure information of garment fastly and accurately from the clothing images is the key of 3D garment modeling. We apply the improved graph cut algorithm to realize the clothing image interactive segmentation, and extract the contour feature of garment. We use the shape matching algorithm based on dynamic programming to merge the feature and style template partitionly, and extract the structure information of the clothing parts.We propose a modeling template construction method based on step fusion of the features and the human template. Different style of modeling template is the foundation of the method of the 3D garment modeling. On the basis of the human template, according to the characteristics of the clothes with different styles, we mix the loose quantity characteristics and the human template together, and realize the generation of the clothing geometry structure; with the fusion of the style features and the human template, we realize the construction of the template topology structure; with the fusion of the size features and the human template, we adopt different construction method of the different parts on modeling templates, and realise the modeling template with different styles of clothing structure.We propose a 3D garment modeling method based on the fusion of features and garment modeling template. The integration of various characteristics and modeling template effectively is the guarantee of the 3D garment modeling. According to the styles information of clothing in the image, we select the corresponding modeling template. With the fusion of the image shape features and modeling template, we use contour matching technology to driven the surface mesh of modeling template deformation; with the fusion of the size features and the modeling template, we use multi target optimization algorithm to control the size problem of 3D garment. Finally,We integrate the image texture feature and the modeling template, realize the texture mapping by using two step texture mapping technology, and obtain an realistic 3D garment modeling.We apply the research work of this paper as the core technology, realize an image-based of 3D garment modeling prototype system, and integrate it into the 3D garment design software named LookStailorX, and then give some case study of the aboved work. Finally, we summarize the main content of the whole work, address some problems existing in the research, and put forward some prospects of future work.