Real-time 3D data reduction and reproduction of spatial model using line detection in RGB image

We propose a real-time polygon reduction method for online reproduction of a three-dimensional spatial model using image processing. Currently, Microsoft Kinect is a popular device for capturing a wide-area and detailed depth image. When producing 3D data reconstruction from the depth image and the RGB image as well as when transmitting 3D data in realtime, it is important to reduce the data size. In this paper, we introduce a polygon reduction method that uses line detection for architectural surfaces and their joint lines in an RGB image. The system discards most of the depth information, leaving the representative value of the surface with part of the RGB image. The remaining data are used to reconstruct simple polygon data drawn by point-cloud or texture mappings. From the operation test of our proposed method, it is confirmed that the polygon reduction could reduce data without increasing the time duration of 3D reconstruction.