从单幅图像恢复立体景象的新方法

Novel approach of recovering stereo view scene from a single image

在研究单幅平面图像内在特性的基础上,提出了一种恢复立体视觉景象建模的新方法。对图像进行智能识别处理,可以求得许多线段的特征参数,并由此计算出消隐点和消隐线,从而可自动获得场景的立体结构信息。本算法的特点在于用一个代数表达式统一了三种典型的度量方法,无需传统的相机内校正参数,直接可计算出建模用立体信息。建模结果用VRML格式保存、输出,以便于网上浏览。众多的图像验证了该方法的有效性、适用性。

A novel approach for reconstructing texture-mapped, 3D scene models from a single painting or photograph is presented based on inside characteristics of single view. Gained a sparse set of characteristic parameters in three directions by automatic recognition, the vanishing points and vanishing lines are computed. With these projective attributes, the approach can build the model automatically. The key feature of the approach is an algebraic representation which unifies three typical measurements, and the approach keeps users from the problem of the camera's internal calibration. We can save the data with VRML format, which gives a convenient way for browsing on internet. The approach has been shown to yield high quality results on a large variety of images.