Proceedings of the 27th annual conference on Computer graphics and interactive techniques
Quasi-Dense Reconstruction from Image Sequence
ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part II
Marching through the Visible Man
VIS '95 Proceedings of the 6th conference on Visualization '95
Distinctive Image Features from Scale-Invariant Keypoints
International Journal of Computer Vision
Unsupervised 3D Object Recognition and Reconstruction in Unordered Datasets
3DIM '05 Proceedings of the Fifth International Conference on 3-D Digital Imaging and Modeling
ACM SIGGRAPH 2006 Papers
Photo tourism: exploring photo collections in 3D
ACM SIGGRAPH 2006 Papers
Poisson surface reconstruction
SGP '06 Proceedings of the fourth Eurographics symposium on Geometry processing
3D-modeling by ortho-image generation from image sequences
ACM SIGGRAPH 2008 papers
Efficient Polyhedral Modeling from Silhouettes
IEEE Transactions on Pattern Analysis and Machine Intelligence
Volumetric Descriptions of Objects from Multiple Views
IEEE Transactions on Pattern Analysis and Machine Intelligence
| Hi-index | 0.00 |
This paper explores the practical aspects associated with visual-geometric reconstruction of a complex 3D scene from a sequence of unconstrained and uncalibrated 2D images. These image sequences can be acquired by a video camera or a handheld digital camera without the need for camera calibration. Once supplied with the input images, our system automatically processes and produces a 3D model. We propose a novel approach, which integrates uncalibrated Structure from Motion (SfM), shape-from-silhouette and shape-from-correspondence, to create a quasi-dense scene geometry of the observed scene. In the second stage, surface and texture are applied onto the generated scene geometry to produce the final 3D model. The advantage of combining silhouette-based and correspondence-based reconstruction approaches is that the new hybrid system is able to deal with both featureless objects and objects with concaved regions. These classes of objects usually pose great difficulty for shape-from-correspondence and shape-from-silhouette approach. As the result, our approach is capable of producing satisfactory results for a large class of objects. Our approach does not require any a priori information about camera and image acquisition parameters. We tested our algorithm using a variety of datasets of objects with different scales, textures and shapes acquired under different lighting conditions. The results indicate that our algorithm is stable and enables inexperienced users to easily create complex 3D content using a standard consumer level camera.