Automatic Depth Map Estimation of Monocular Indoor Environments

3D-TV is well expected as next advancement in television. The conversion technique of two-dimensional (2D) images to 3D images is an effective way to alleviate the shortage of 3D contents. The most important and difficult problem in 2D-to-3D conversion is how to generate or estimate depth information using only a single-view image. In this paper, a new approach to estimate depth map from monocular indoor environments is presented. Our main contribution consists of two expects, one is to introduce the concept of planar model to 2D-to-3D conversion technique, and the other is the methodology to extract planes and successively estimate 2D image depth. In our depth estimation algorithm, indoor image is firstly partitioned into foreground and background, then background is interpreted in terms of one or two horizontal and several vertical planes. Finally, the depth map of image is obtained according to the plane classification machine and geometric feature. Examples of depth generation and 3D effects are presented and discussed.