Vision Modeling and Visualization · 2007 · 15 citations · 12 references
ComputingComputer ScienceComputer VisionDigital ImagingPhotographyCamera MovementCamera ProjectionDigital PhotographyImage RegistrationMotion CorrectionMotion EstimationCamerasImage Alignment For HdrMotion DetectionParallax CorrectionParallax EstimationTransformations
In theory HDR photography generated from expo-sure sequences is limited to pictures of still scenes preferably taken with the camera mounted on a tri-pod. Current state-of-the-art HDR imaging soft-ware relaxes this limitation by aligning the pictures with afne transformations. However, two pictures cannot be aligned with a single afne transforma-tion if objects in the scene move or the position of the camera changes which causes parallax effects in the image. In this paper, a method for the non-linear align-ment of a picture sequence for HDR imaging is pre-sented. By using techniques for motion detection and compensation, the approach is not only capa-ble to correctly align pictures containing parallax effects, but also able to compensate for the move-ment of objects in the scene. This way, taking free-hand HDR pictures from non-static scenes becomes possible with off-the-shelf digital cameras. 1
12
Recovering high dynamic range radiance maps from photographs
Paul Debevec, Jitendra Malik · 2008 · 2.4K citations
Gradient domain high dynamic range compression
Raanan Fattal, Dani Lischinski, Michael Werman · 2002 · 1.1K citations
Performance of optical flow techniques
John A. Barron, David J. Fleet, Steven S. Beauchemin et al. · 2003 · 740 citations
Computing, Physics, Computer Vision +10