Co-Analysis and Parameterization of 3D Shape Collections for Shape Synthesis (DAEV)

With online model repositories growing larger every day, both experienced and inexperienced modelers are presented with new possibilities for content creation. One such possibility is the creation of new shapes by combining parts of already existing shapes. The advantages of this shape synthesis method are that it takes less time than traditional modeling approaches and that it can be used even by inexperienced users. This thesis introduces a framework for this type of shape synthesis that consists of four stages, incorporating a new way for parameterization and exploration of shape collections. Using a modular and extensible approach, the co-analysis stage groups parts of shapes into categories based on their function, creating a correspondence between parts of different shapes. By analyzing relations between pairs of parts and how their spatial arrangements vary across the collection, a small number of parameters is found in the parameterization stage. Starting with an initial shape, these parameters can then be used to browse the collection in the exploration stage, either by altering the parameters directly or by interacting with the shape itself. Finally, in the synthesis stage a new shape can be created by exchanging parts of the initial shape with corresponding parts of the shapes found during the exploration.