Anisotropic Huber-L1 Optical Flow

By Manuel Werlberger, Werner Trobin, Thomas Pock, Andreas Wedel, Daniel Cremers, Horst Bischof. British Machine Vision Conference, 2009

448 citations recorded by the source corpus.

Subfields: Computational Theory and Mathematics, Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/65635c3a61b979625e44c986ac281c35ac3b0667

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

Optical Flow Estimation Using a Spatial Pyramid Network (2016, methodology): “Using a GPU can speed up traditional methods [37, 43] but with reduced accuracy.”
EpicFlow: Edge-preserving interpolation of correspondences for optical flow (2015, background): “Effective approaches were previously proposed for handling the case of small displacements ( i.e ., less than a few pixels) [19, 35, 24].”
Dense Semi-rigid Scene Flow Estimation from RGBD Images (2014, background): “Moreover, the L(1) norm can be replaced by the Huber norm [23] to reduce the staircasing effect.”
Optical flow modeling and computation: A survey (2015, methodology): “For example, the results reported in [233,244] are obtained with an initialization by the result of the methods of [231,271] respectively.”
Efficient sparse-to-dense optical flow estimation using a learned basis and layers (2015, methodology): “For each movie, we compute the optical flow using GPUFlow [52].”
Joint motion estimation and segmentation of complex scenes with label costs and occlusion modeling (2012, methodology): “Results are shown on several classical motion segmentation and optical flow examples.”
New features and insights for pedestrian detection (2010, methodology): “2In our previous paper [27] we used the optic flow software of [29], which is a precursor of [26].”
3D Scene Flow Estimation with a Piecewise Rigid Scene Model (2015, methodology): “To deﬁne these weights we follow Werlberger et al. (2009) and apply the anisotropic diffusion tensor: The image gradient direction g = ∇ I / |∇ I | is determined via bicubic interpolation in the middle between p and q .”
Layered RGBD scene flow estimation (2015, background): “Some notable variants include image-dependent, non-local motion priors to recover sharp motion boundaries [38, 47, 51], fusion moves to reach stronger local minima [24], embedding feature matching to track small, fast-moving objects [10, 45, 49], exploiting the epipolar constraint for egocentric sce”
Dense Reconstruction Using 3D Object Shape Priors (2013, background): “d(u) is the depth map gradient, γ is a scalar weighting the effect of the regularization over the photo-consistency and ‖·‖ is the Huber norm of the intensity difference [25].”
Phase-based frame interpolation for video (2015, unknown): “(1) Let us now modify the phase difference according to a weight α ∈ (0, 1) that describes an intermediate position between the functions: y = A sin(ωx−αφdiff ) = A sin(ω(x−αφshift)) .”
Fast Optical Flow Using Dense Inverse Search (2016, methodology): “Robust error norms , such as L 1 and the Huber-norm[52], can be used instead of the L 2 -norm, implicit in the optimization of (1).”
A survey of appearance models in visual object tracking (2013, unknown): “…2007] Global Matrix-based raw pixel representation 3 [Wang et al. 2007] Global Multi-cue raw pixel representation (i.e., color, position, edge) 4 [Werlberger et al. 2009; Santner et al. 2010] Global Optical flow representation (constantbrightness-constraint) 5 [Black and Anandan 1996; Wu and Fan…”
Nonlinear disparity mapping for stereoscopic 3D (2010, methodology): “For video with moving scene elements one has to ensure that local image distortion is properly transferred along the local motion flow [Werlberger et al. 2009] between successive video frames.”
Spatio-temporal video autoencoder with differentiable memory (2015, methodology): “We use Huber loss as penalty, with its corresponding derivative (2), due to its edge-preserving capability Werlberger et al. (2009).”
Layered segmentation and optical flow estimation over time (2012, methodology): “HGVS uses the output from a recent optical flow estimation method [34].”
A Database and Evaluation Methodology for Optical Flow (2007, methodology): “For example, Nagel and Enkelmann (1986) and Werlberger et al. (2009) weight the direction along the image gradient less than the direction orthogonal to it, and Sun et al. (2008) learn a Steerable Random Field to define the weighting.”