|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Home | Research | Research Programs | Computational Video |
Computational VideoSurface Reconstruction from Multiple Images
Surface reconstruction is concerned with the problem of creating geometric models from multiple images. Such models are important in a wide variety of applications, including virtual/augmented reality, reverse engineering and animation. However, when light rays from 3D objects project onto 2D planes to form images, a great deal of information is lost, which leads to ill-posed mathematical problems. In general, solutions that are unique and stable can only be guaranteed with additional constraints. Previous solutions to this problem have relied on strong assumptions made about the underlying surface, such as the planar or quadric surface. In this project, researchers instead make minimal assumptions and seek more general solutions by exploiting the intrinsic geometric constraints within the multi-view images. Using projective vision techniques, researchers compute camera parameters and positions from uncalibrated image sequences. By assuming the local smoothness of the underlying surfaces, they construct compact representations - such as B-splines - from very few feature points. Researchers study robust statistical and numerical methods involved in the process of reconstruction. They also develop statistical models for various estimation problems in order to evaluate the accuracy and reliability of the reconstructed surfaces. Ultimately, researchers expect to develop efficient algorithms for surface reconstruction with minimal user assists. Research ContactDr. Chang Shu Business ContactDr. George Forester |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|