Universität Des Saarlandes Fachrichtung 6.1 – Mathematik Lucas/kanade Meets Horn/schunck: Combining Local and Global Optic Flow Methods Lucas/kanade Meets Horn/schunck: Combining Local and Global…

By Joachim Weickert, A. Bruhn, Christoph Schnörr

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Generative Visual Manipulation on the Natural Image Manifold (2016, methodology): “Motion+Color flow algorithm: We then estimate the color and geometric changes by generalizing the brightness constancy assumption in traditional optical flow methods [30,31].”
SIFT Flow: Dense Correspondence across Scenes and Its Applications (2011, background): “Image alignment, a.k.a. image registration or correspondence, is a broad topic in computer vision, computer graphics and medical imaging, covering stereo, motion analysis, video compression, shape registration, and object recognition.”
Unsupervised convolutional neural networks for motion estimation (2016, methodology): “In order to deal with motions large in magnitude, we embed the proposed network in a classical iterative scheme, in which at the end of each iteration the reference image is warped towards the target image and in a classical coarse-to-fine multiscale framework.”
Fast and Accurate Image Super-Resolution with Deep Laplacian Pyramid Networks (2017, background): “where rðxÞ 1⁄4 ffiffiffiffiffiffiffiffiffiffiffiffiffiffi x2 þ 2 p is the Charbonnier penalty function (a differentiable variant of L1 norm) [42], N is the number of training samples in each batch, S is the target upsampling scale factor, and L 1⁄4 log 2 S is the number of pyramid levels in our mode”
Scale-Space Flow for End-to-End Optimized Video Compression (2020, background): “[6, 18, 13]) as well as the use of multi-scale pyramids in deep flow estimation networks [28, 14].”
A primal-dual framework for real-time dense RGB-D scene flow (2015, methodology): “The operators ∂ and ∂− represent right and left derivatives and are approximated using the stencils [0 -1 1] and [-1 1 0] respectively, and the terms r x , r − x , r + y , r − y are right and left approximations of the ones presented in (15).”
Distributed cosegmentation via submodular optimization on anisotropic diffusion (2011, result): “As compared in Table 1, our approach has some unique benefits including compelling performance over previous methods, superior scalability, and a desirable ability of automatically deciding the number of segments.”
Deformable Model Fitting by Regularized Landmark Mean-Shift (2010, methodology): “…model, such as that in (1), the utility of such a framework has been demonstrated using more generic constraints in problems such as optical flow (Bruhn et al. 2005) and stereo matching (Sun et al. 2003), where constraints take the form of a smoothing process over the motion domain of landmarks.”
PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume (2017, methodology): “As it is computationally impractical to perform a full search, a coarse-tofine, warping-based approach is often adopted [11].”
Beyond pixels: exploring new representations and applications for motion analysis (2009, methodology): “Techniques such as iterative reweighted least squared (IRLS) [22, 20], pyramid-based coarse-to-fine search [14, 22] and occlusion/outlier detection [95] were intensively used for optimizing these nonlinear objective functions.”
Global Optimization for Shape Fitting (2007, methodology): “Then, they use some heuristics deviating from (3) and causing geometric artifacts (e.g. “thickness”).”
CoTracker: It is Better to Track Together (2023, background): “Originally approached by studying the colour constancy equation [3, 4, 17, 27], optical flow is nowadays tack-led by using deep learning, starting with FlowNet [12, 19].”
DF-Net: Unsupervised Joint Learning of Depth and Flow using Cross-Task Consistency (2018, methodology): “Leveraging the brightness constancy and spatial smoothness priors used in classical dense correspondence algorithms [4,23,40], prior work has used the photometric discrepancy between the warped frame and the target frame as an unsupervised proxy loss function for training CNNs without ground truth a”
Motion-based counter-measures to photo attacks in face recognition (2014, methodology): “The core of the algorithm in this implementation is based on [24] and [25].”