A 3D Skeleton Points-Based Hierarchical Body Modeling Approach for Intelligent Online Clothing Fitting Systems

The online clothing fitting (OCF) has been a prevalent Internet application. It utilizes multimedia processing algorithms to generate clothing fitting effect images for consumers in virtual way. Existing related works mainly neglected the diverse skeleton structure characteristics of different users, which limits generalization of OCF systems. To handle such challenge, this paper proposes a novel 3D skeleton points-based hierarchical body modeling approach for intelligent OCF systems. Specifically, it focuses on two types of posture features, integrates circumferencial features of body, and distinguishes the trunks and the limbs. Then, a global-local singular value decomposition algorithm is designed to hierarchically control the deformation of 3D skeleton format cooperatively, in order to optimize the fitting results. Finally, we conduct some experimental evaluation through computer programming to verify efficiency of the proposal. The results show that relative tracking error of the new method is reduced by about 10%, and the real-time tracking accuracy between the clothing fitting results and the human body is obviously improved.