3D spatial layout extraction of indoor images using RGB-D data

In this paper we present a new method for 3d spatial layout extraction of indoor images. First, we get room orientation and major planes by using color and depth data. Then, we find potential wall planes that form the room boundaries and evaluate them with geometric room hypothesises. In the experiments with dataset which is collected using Microsoft Kinect sensor, we observe that our approach is successful to extract room wall surfaces and corner edges. The experimental results show that the proposed method can improve the performance of image analysis stages by decreasing the complexity of image and the problem space.