Computing visual correspondence with occlusions using graph cuts

By Vladimir Kolmogorov, Ramin Zabih. Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, 2001

1,375 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/22d606ae7351aa4e9598ed6194797d03557abe12

How researchers see this paper

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

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Joint motion estimation and segmentation of complex scenes with label costs and occlusion modeling (2012, background): “In [10], a map uniqueness constraint is embedded in a graph cut framework for stereo and motion estimation.”
Advances in Computational Stereo (2003, background): “Zabih and Woodfill [89] propose a variation of the rank transform, called the census transform, that preserves the spatial distribution of ranks by encoding them in a bit string (see Fig.”
Continuous Markov Random Fields for Robust Stereo Estimation (2012, neutral): “> 2 pixels > 3 pixels > 4 pixels > 5 pixels Non-Occ All Non-Occ All Non-Occ All Non-Occ All GC+occ [36] 39.”
SIFT Flow: Dense Correspondence across Scenes and Its Applications (2011, background): “The chance that some of the nearest neighbors share the same scene characteristics with a query image increases as the database grows, and the correspondence obtained by SIFT flow can be semantically meaningful.”
Global stereo reconstruction under second order smoothness priors (2008, methodology): “used [33], [34], but this enforces the ordering constraint, a constraint which is often violated in real scenes, especially those with thin foreground objects [35]; some formulations without the ordering constraint can be optimized using α-expansion graph cuts [11], [15], [17], [28], but it has been”
Stereo Matching with Color-Weighted Correlation, Hierachical Belief Propagation and Occlusion Handling (2006, background): “As mentioned earlier there typically a data term and smoothness term.”
CUDA cuts: Fast graph cuts on the GPU (2008, methodology): “Graph-cuts have since then been applied to several Computer Vision problems like image and video segmentation [29, 27], stereo and motion [5, 31], multicamera scene reconstruction [24, 20], etc.”
A fast area-based stereo matching algorithm (2004, background): “[14], rely on iterative schemes that carry out disparity assignments on the basis of the minimisation of a global cost function.”
Efficient Large-Scale Stereo Matching (2010, methodology): “These parameters are set based on suggestions in [5] and adaptions to the largescale imagery we use.”
HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching (2020, methodology): “Traditional methods utilize handcrafted schemes to find local correspondences [55, 22, 4, 21] and global optimization to exploit spatial context [14, 26, 27].”
Learning Conditional Random Fields for Stereo (2007, background): “In particular, MRF models for stereo have become popular since high-quality approximate solutions can be obtained efﬁciently using graph cuts (GC) [6–8] and belief propagation (BP) [9–11].”
An experimental comparison of min-cut/max- flow algorithms for energy minimization in vision (2001, methodology): “The corresponding graph structures were previously described by [15, 18, 4, 8, 23, 24, 5] in detail.”
Group-Wise Correlation Stereo Network (2019, methodology): “In contrast, global methods minimize a target function to solve the optimal disparity map, which usually takes both matching costs and smoothness priors into consideration, such as belief propagation [23, 9] and graph cut [11].”
Stereo Matching Using Belief Propagation (2002, methodology): “However, the uniqueness constraint is hard to impose as in [ 22 ] because we use a probabilistic distribution as the final solution.”
AANet: Adaptive Aggregation Network for Efficient Stereo Matching (2020, methodology): “Global methods usually solve an optimization problem by minimizing a global objective function that contains data and smoothness terms [31, 17], while local methods only consider neighbor information [40, 12], making themselves much faster than global methods [23, 29].”
Ieee Transactions on Pattern Analysis and Machine Intelligence 1 Stereo Processing by Semi-global Matching and Mutual Information (methodology): “Consequently, most of these methods have runtimes of more than 20 seconds [3], [12] to more than a minute [9]–[11], [13], [14] on the test images.”