3D shape recovery from single image by using texture information

In this paper, we propose a three-dimension shape recovery method that works with a single image without prior information about an image, a camera or a scene. Three-dimension shape recovery can be produced from the depth map that is generated from the convolution between the 2D planar image and the texture gradient image. The experimental results have been presented demonstrating the 3D imagepsilas surfaces and the comparisons of cross-section images between the 3D model from the digitizer and our proposed method.