Robust L/sub 1/ norm factorization in the presence of outliers and missing data by alternative convex programming

  title={Robust L/sub 1/ norm factorization in the presence of outliers and missing data by alternative convex programming},
  author={Qifa Ke and Takeo Kanade},
  journal={2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05)},
  pages={739-746 vol. 1}
  • Qifa KeT. Kanade
  • Published 20 June 2005
  • Computer Science
  • 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05)
Matrix factorization has many applications in computer vision. Singular value decomposition (SVD) is the standard algorithm for factorization. When there are outliers and missing data, which often happen in real measurements, SVD is no longer applicable. For robustness iteratively re-weighted least squares (IRLS) is often used for factorization by assigning a weight to each element in the measurements. Because it uses L/sub 2/ norm, good initialization in IRLS is critical for success, but is… 

Figures from this paper

A Cyclic Weighted Median Method for L1 Low-Rank Matrix Factorization with Missing Entries

This paper proposes a novel cyclic weighted median (CWM) method, which is intrinsically a coordinate decent algorithm, for L1-norm LRMF, and demonstrates that the CWM method outperforms state-of-the-arts in terms of both accuracy and computational efficiency.

Efficient computation of robust low-rank matrix approximations in the presence of missing data using the L1 norm

This paper presents a method for calculating the low-rank factorization of a matrix which minimizes the L1 norm in the presence of missing data and shows that the proposed algorithm can be efficiently implemented using existing optimization software.

Divide-and-Conquer Method for L1 Norm Matrix Factorization in the Presence of Outliers and Missing Data

It is substantiated that the divide-and-conquer method, proposed, achieves better results than the current state-of-the-art methods on $L1$ matrix factorization calculation in both computational time and accuracy, especially on large-scale applications such as face recognition and structure from motion.

Robust Matrix Factorization with Unknown Noise

A low-rank matrix factorization problem with a Mixture of Gaussians (MoG) noise, which is a universal approximator for any continuous distribution, and hence is able to model a wider range of real noise distributions.

Practical low-rank matrix approximation under robust L1-norm

This work proposes to add a convex trace-norm regularization term to improve convergence, without introducing too much heterogenous information, and customize a scalable first-order optimization algorithm to solve the regularized formulation on the basis of the augmented Lagrange multiplier (ALM) method.

Smooth incomplete matrix factorization and its applications in image/video denoising

On the minimal problems of low-rank matrix factorization

This paper proposes a new parametrization and a building-block scheme to solve low-rank matrix factorization by extending the solution from a small sized minimal problem to generate all the minimal problems of a given size using Laman graph theory.

Large-Scale Matrix Factorization with Missing Data under Additional Constraints

The empirical evaluations suggest that, under the conditions of matrix completion theory, the proposed algorithm finds the optimal solution, and also requires fewer observations compared to the current state-of-the-art algorithms.

A Probabilistic Approach to Robust Matrix Factorization

Experiments show that PRMF is comparable to other state-of-the-art robust matrix factorization methods in terms of accuracy and outperforms them particularly for large data matrices.



Factorization with Uncertainty

A new approach to covariance-weighted factorization, which can factor noisy feature correspondences with high degree of directional uncertainty into structure and motion and provides a unified approach for treating corner-like points together with points along linear structures in the image.

Factorization with Uncertainty and Missing Data: Exploiting Temporal Coherence

This paper uses the well known EM algorithm for factor analysis to perform factorization and places a prior on the temporal trajectory of the latent variables (the camera position) and shows that incorporating this prior gives a significant improvement in performance in challenging image sequences.

Robust Factorization

A new and computationally efficient algorithm for applying an arbitrary error function in the factorization scheme that enables the use of robust statistical techniques and arbitrary noise models for the individual features.

A Framework for Robust Subspace Learning

The theory of Robust Subspace Learning (RSL) for linear models within a continuous optimization framework based on robust M-estimation is developed and applies to a variety of linear learning problems in computer vision including eigen-analysis and structure from motion.

Linear Fitting with Missing Data for Structure-from-Motion

  • D. Jacobs
  • Computer Science
    Comput. Vis. Image Underst.
  • 2001
This work proposes a novel method for fitting a low rank matrix to a matrix with missing elements and shows that this method has desirable theoretical properties compared to previously proposed methods, because it can find solutions when there is less data present.

Outlier correction in image sequences for the affine camera

This work presents an outlier correction scheme that iteratively updates the elements of the image measurement matrix, with the result that outliers are corrected and retained, giving improved reconstruction and smaller reprojection errors.

Shape and motion from image streams under orthography: a factorization method

A factorization method is developed that can overcome the difficulty by recovering shape and motion under orthography without computing depth as an intermediate step, and gives accurate results.

Learning the parts of objects by non-negative matrix factorization

An algorithm for non-negative matrix factorization is demonstrated that is able to learn parts of faces and semantic features of text and is in contrast to other methods that learn holistic, not parts-based, representations.

Multi-frame optical flow estimation using subspace constraints

  • M. Irani
  • Engineering, Computer Science
    Proceedings of the Seventh IEEE International Conference on Computer Vision
  • 1999
This paper develops a method for simultaneous estimation of optical flow across multiple frames, which uses multi-frame subspace constraints to constrain the 2D correspondence estimation process itself, and not for 3D recovery.

On the unification of line processes, outlier rejection, and robust statistics with applications in early vision

It is shown how prior assumptions about the spatial structure of outliers can be expressed as constraints on the recovered analog outlier processes and how traditional continuation methods can be extended to the explicit outlier-process formulation.