3D Shape Acquisition and Arbitrary View Image Generation from Monocular Image Based on Primitive Decomposition

It is difficult to reconstruct 3D shape of an object from information of a monocular image. In this paper, the authors propose a method for 3D shape acquisition and arbitrary view image generation of objects on the ground. By using 3D spatial information of background and dividing object´s shape into simple shape such as rectangular parallelepiped and triangle pole, the authors achieve to acquire it from a monocular image. In some experimental results, the authors show that 3D shape can be simply estimated and arbitrary view images can be generated from monocular vision