Automatic Textures Updating Method for 3D Indoor Scenes Based on Mobile Phone Images

Nowadays, the texture mapping for 3D indoor model mostly relies on heavy artificial operation. Its inefficiency and difficulty in realizing texture updating after the change of indoor environment leads to a lot of out-dated and unreal texture in time. To solve this problem, we propose a new method for updating 3D indoor model texture automatically. Using a single realistic image from smartphone, we firstly estimate the spatial layout of the room in the image based on 3D indoor scene understanding. Then, according to phone's direction and position, we find the triangular patches of the targeted wall in the 3D geometric model. At the end, the (u, v) coordinates of the image are calculated by matching the convex points of the layout box with the corresponding patches of the 3D model. The conducted experiments demonstrate that the proposed method can automatically map a single image onto the 3D indoor model, which is meaningful for the augmented reality in the indoor space.