“Locally Reparameterized Light Fields” by Tong, Shum, Feng, Kang and Szeliski

  • ©Xin Tong, Heung-Yeung Shum, Tao Feng, Sing Bing Kang, and Richard Szeliski



Interest Area:



    Locally Reparameterized Light Fields

Session/Category Title:   Light Fields



    In this sketch, we propose a novel IBR representation to handle non-diffuse effects compactly, which we call locally reparameterized light field (LRLF). LRLF is based on the use of local and separate diffuse and non-diffuse geometries. The diffuse geometry is associated with true depth while the non-diffuse geometry has virtual depth that provides local photo-consistency with respect to its neighbors.


    1. Bolles, R.C., Baker, H.H., & Marimont, D.H. (1987). Epipolar-plane image analysis: An approach to determining structure from motion. International Journal of Computer Vision, 1, 7-55.
    2. Gortler, S.J., Grzeszczuk, R., Szeliski, R., & Cohen, M.F. (1996). The lumigraph. Proceedings of SIGGRAPH 96, 43-54.
    3. Szeliski, R., Avidan, S., & Anandan, R. (2000). Layer extraction from multiple images containing reflections and transparency. In Conference on Computer Vision and Pattern Recognition, I, 246-253.

Overview Page: