Robust Fitting in Computer Vision: Easy or Hard?

@inproceedings{Chin2018RobustFI,
  title={Robust Fitting in Computer Vision: Easy or Hard?},
  author={Tat-Jun Chin and Zhipeng Cai and Frank Neumann},
  booktitle={ECCV},
  year={2018}
}
Robust model fitting plays a vital role in computer vision, and research into algorithms for robust fitting continues to be active. Arguably the most popular paradigm for robust fitting in computer vision is consensus maximisation, which strives to find the model parameters that maximise the number of inliers. Despite the significant developments in algorithms for consensus maximisation, there has been a lack of fundamental analysis of the problem in the computer vision literature. In… Expand
Robust Fitting in Computer Vision: Easy or Hard?
TLDR
This work presents several computational hardness results for consensus maximisation, underline the fundamental intractability of the problem, and resolve several ambiguities existing in the literature. Expand
GORFLM: Globally Optimal Robust Fitting for Linear Model
TLDR
A deterministic globally optimal linear model fitting method is proposed, in which the negative Gaussian function is used as a soft loss function over the residual and formulates model fitting as minimizing the sum of the Gaussian functions. Expand
Unsupervised Learning for Robust Fitting: A Reinforcement Learning Approach
TLDR
This paper introduces a novel unsupervised learning framework that learns to directly solve robust model fitting, and achieves competitive results compared to traditional methods on several important computer vision problems. Expand
Model-free Consensus Maximization for Non-Rigid Shapes
TLDR
This paper forms the model-free consensus maximization as an Integer Program in a graph using ‘rules’ on measurements and provides a method to solve it optimally using the Branch and Bound (BnB) paradigm. Expand
Efficient Algorithms for Maximum Consensus Robust Fitting
TLDR
This article develops two highly efficient deterministic algorithms based on the alternating direction method of multipliers (ADMM) and proximal block coordinate descent (BCD) frameworks that are guaranteed convergent and establish some meaningful properties in the BCD algorithm. Expand
Consensus Maximization Tree Search Revisited
TLDR
It is shown that the consensus maximization tree structure used previously actually contains paths that connect nodes at both adjacent and non-adjacent levels, and a new acceleration strategy is proposed that avoids such redundant paths. Expand
Special Issue: Advances in Architectures and Theories for Computer Vision
TLDR
This special issue focuses on advances in architectures and theories for computer vision, a community that continues to adapt to new tools, but also contributes to the development of the tools in a way that is informed by the specific challenges of computer vision. Expand
A Novel Method for the Absolute Pose Problem with Pairwise Constraints
TLDR
This paper considers pairwise constraints and proposes a globally optimal algorithm for solving the absolute pose estimation problem and has a linear complexity in the number of correspondences at a given outlier ratio. Expand
Convex Relaxations for Consensus and Non-Minimal Problems in 3D Vision
TLDR
The main contribution of this paper is the claim that a good approximate solution for many polynomial problems involved in 3D vision can be obtained using the existing theory of numerical computational algebra. Expand
Deterministic consensus maximization with biconvex programming
TLDR
This paper proposes an efficient deterministic optimization algorithm for consensus maximization that conducts a deterministic search that forcibly increases the consensus of the initial solution. Expand
...
1
2
3
4
...

References

SHOWING 1-10 OF 41 REFERENCES
Deterministically maximizing feasible subsystem for robust model fitting with unit norm constraint
TLDR
This work proposes a deterministic branch and bound method to solve the MaxFS problem with guaranteed global optimality and introduces a piecewise linear relaxation to build very tight under- and over-estimators for square terms by partitioning variable bounds into smaller segments. Expand
Robust Fitting for Multiple View Geometry
TLDR
It is shown that for most fitting problems, a solution can be found with an algorithm that has polynomial time-complexity in the number of points (independent of the rate of outliers), and other cost functions can also be handled within the same framework with the same time complexity. Expand
Efficient globally optimal consensus maximisation with tree search
TLDR
It is shown how consensus maximisation for a wide variety of vision tasks can be posed as a tree search problem, which leads to a novel algorithm based on A* search that is several orders of magnitude faster than previous exact methods. Expand
Guaranteed Outlier Removal with Mixed Integer Linear Programs
TLDR
This work proposes an algorithm based on mixed integer linear programming to perform guaranteed outlier removal as a technique to reduce the runtime of exact algorithms, and demonstrates that overall speedups of up to 80% can be achieved on common vision problems. Expand
A Branch-and-Bound Approach to Correspondence and Grouping Problems
TLDR
This paper presents a new algorithm that combines the benefits of both appearance-based and geometry-based methods and mathematically guarantees a global optimization and generalizes it in the context of data correspondence/grouping under an unknown parametric model and shows it can be applied to certain classes of computer vision problems. Expand
Guaranteed Outlier Removal for Rotation Search
TLDR
This paper proposes a novel outlier removal technique for rotation search that guarantees that any correspondence it discards as an outlier does not exist in the inlier set of the globally optimal rotation for the original data. Expand
Maximum Consensus Parameter Estimation by Reweighted ℓ1 Methods
TLDR
This work proposes a smooth surrogate function, the minimization of which leads to an extremely simple iteratively reweighted algorithm for MaxCon that is very efficient and in many cases, yields the global solution. Expand
An Exact Penalty Method for Locally Convergent Maximum Consensus
  • Huu Le, Tat-Jun Chin, D. Suter
  • Computer Science, Mathematics
  • 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • 2017
TLDR
This paper develops a Frank-Wolfe algorithm that can deterministically solve the maximum consensus problem and is much more practical on realistic input sizes. Expand
Tractable Algorithms for Robust Model Estimation
TLDR
The number of outliers can be minimized in polynomial time with respect to the number of measurements, although exponential in the model dimension, and it is proved that the statistically more desirable truncated $$L_2$$L2-norm can be optimized with the same complexity. Expand
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography
TLDR
New results are derived on the minimum number of landmarks needed to obtain a solution, and algorithms are presented for computing these minimum-landmark solutions in closed form that provide the basis for an automatic system that can solve the Location Determination Problem under difficult viewing. Expand
...
1
2
3
4
5
...