HDR & IBR

    Friday, 22 November

    09:00 - 10:45

    Room S224 + S225

    Image-Based Rendering in the Gradient Domain

    We propose a novel gradient domain image-based rendering algorithm for handling complex scenes that may include reflective surfaces. We estimate the depths of image gradients rather than pixels, and obtain a final result in real-time through GPU accellerated Poisson integration using an approximate solution as a data term.

    Johannes Kopf, Microsoft Research
    Fabian Langguth, Technische Universitaet Darmstadt
    Daniel Scharstein, Middlebury College
    Richard Szeliski, Microsoft Research
    Michael Goesele, Technische Universitaet Darmstadt

    Data-driven Hallucination of Different Times of Day from a Single Outdoor Photo

    We introduce “time hallucination”: synthesizing a plausible image
    at a different time of day from an input image.

    YiChang Shih, Massachusetts Institute of Technology
    Sylvain Paris, Adobe Research
    Fredo Durand, Massachusetts Institute of Technology
    William T. Freeman, Massachusetts Institute of Technology

    Automatic Noise Modeling for Ghost-free HDR Reconstruction

    Reconstructing HDR images of dynamic scenes from multiple exposures is challenging: Moving objects introduce ghosting artifacts.
    For avoiding these artifacts, we automatically model the image's noise distribution and then estimate the likelihood that corresponding colors be observations of the same object. Cluttered and fast-moving scenes can now be reconstructed.

    Miguel Granados, Max Planck Institute for Informatics
    Kwang In Kim, Max Planck Institute for Informatics
    James Tompkin, Max Planck Institute for Informatics
    Christian Theobalt, Max Planck Institute for Informatics

    Patch-based High Dynamic Range Video

    In this paper, we propose a novel patch-based optimization algorithm for generating High Dynamic Range (HDR) videos using an off-the-shelf camera. The results presented are high-quality and superior to those produced with existing techniques.

    Nima Khademi Kalantari, University of California at Santa Barbara
    Eli Shechtman, Adobe Systems Inc.
    Connelly Barnes, Adobe Systems Inc.
    Soheil Darabi, Adobe Systems Inc.
    Dan B Goldman, Adobe Systems Inc.
    Pradeep Sen, University of California at Santa Barbara