A real-time virtual dressing system with RGB-D camera

As RGB-D cameras become cheaper commodities, many human-computer interaction applications are facilitated, such as human-machine interface (HMI) and virtual reality (VR). In this paper, we focus on virtual dressing, which is a newly heated application of human-computer interaction, and develop a novel framework for real-time virtual dressing. We made three major contributions in our framework: 1) We introduce an effective algorithm to pre-process and improve the skeleton data from an RGB-D camera; 2) We develop a strategy to measure the skeleton motion parameters (scaling, translation and rotation) and apply a two-step deforming scheme specific to create dress deformation results. 3) We introduce hand-integration module to integrate the clothing models and user arms, so as to guarantee user arms being properly displayed in the visual dressing result. Experimental results show that our proposed virtual dressing system achieves satisfactory visual dressing results.