Research: Robust Analysis of Visual Data

Research: Robust Analysis of Visual Data http://coewww.rutgers.edu/riul/research/robust.html

High breakdown point robust estimators in statistics tolerate up to fifty percent of the data points not obeying the same model. In image analysis, however, the data is often complex and several instances of a model are simultaneously present, each accounting for a relative small percentage of the data points. Only robust methods designed with the special nature of the visual data in mind can achieve satisfactory results. Several papers exploit the mean shift technique for nonparametric clustering of multimodal feature spaces. Code  Semi-Supervised Kernel Mean Shift Clustering Matlab code to perform mean shift clustering in kernel space by using a few user-specified pairwise constraints. The theory is described in Semi-Supervised Kernel Mean Shift Clustering.  For comments, please contact Saket Anand or Sushil Mittal.  Generalized Projection based M-estimator C++ code to find the robust estimate derived without using any user supplied scale. The theory is described in Generalized Projection Based M-Estimator.  For comments, please contact Sushil Mittal or Saket Anand.  Nonlinear Mean Shift over Riemannian Manifolds C++ code to generalize nonlinear mean shift to data points lying on Riemannian manifolds. The theory is described in Nonlinear Mean Shift over Riemannian Manifolds.  For comments, please contact Raghav Subbarao or Sushil Mittal. Edge Detection and Image SegmentatiON (EDISON) System C++ code, can be used through a graphical interface or command line. The system is described in Synergism in low level vision. For comments, please contact Bogdan Georgescu or Chris M. Christoudias.  The EDISON system contains the image segmentation/edge preserving filtering algorithm described in the paper Mean shift: A robust approach toward feature space analysisand the edge detection algorithm described in the paper Edge detection with embedded confidence.  Adaptive mean shift based clustering  C++ code implementing an (approximate) mean shift procedure with variable bandwith (in high dimensions). The algorithm is described in Mean shift based clustering in high dimensions: A texture classification example. For comments, please contact Bogdan Georgescu or Ilan Shimshoni.  Color distribution and optical flow based point matcher C++ code to find point correspondences by matching color distributions computed with spatially oriented kernels and optical flow registration. The theory is described in Point Matching Under Large Image Deformations and Illumination Changes.  For comments, please contact Bogdan Georgescu.  Publications  Please use the link "Abstract" to see the publishing history of a paper. The links "Paper" also contain the abstract. S. Anand, S. Mittal, O. Tuzel, P. Meer: Semi-Supervised Kernel Mean Shift Clustering.  Abstract     Paper (pdf)      Supplementary material.  S. Mittal, S. Anand, P. Meer: Generalized projection based M-Estimator.  Abstract     Paper (pdf)  S. Mittal, P. Meer: Conjugate gradient on Grassmann manifolds for robust subspace estimation.  Abstract     Paper (pdf)  S. Mittal, S. Anand, P. Meer: Generalized projection based M-Estimator: Theory and applications.  Abstract     Paper (pdf)  O. Tuzel, F. Porikli, P. Meer: Kernel Methods for Weakly Supervised Mean Shift Clustering.  Abstract     Paper (pdf)  R. Subbarao, P. Meer: Projection Based M-Estimators.  Abstract    Paper (pdf)  R. Subbarao, P. Meer: Nonlinear Mean Shift over Riemannian Manifolds.  Abstract    Paper (pdf)  R. Subbarao, Y. Genc, P. Meer: Robust Unambiguous Parametrization of the Essential Manifold.  Abstract    Paper (pdf) R. Subbarao, P. Meer: Discontinuity Preserving Filtering over Analytic Manifolds.  Abstract    Paper (pdf) R. Subbarao, Y. Genc, P. Meer: Nonlinear Mean Shift for Robust Pose Estimation.  Abstract    Paper (pdf)  R. Subbarao, P. Meer: Beyond RANSAC: User Independent Robust Regression.  Abstract    Paper (pdf)  R. Subbarao, P. Meer: Nonlinear mean shift for clustering over analytic manifolds.  Abstract    Paper (pdf) R. Subbarao, P. Meer: Subspace estimation using projection based M-estimators over Grassmann manifolds  Abstract    Paper (pdf) O. Tuzel, R. Subbarao, P. Meer: Simultaneous multiple 3D motion estimation via mode finding on Lie groups.  Abstract    Paper (pdf)  Data  O. Tuzel, F. Porikli, P. Meer: A Bayesian approach to background modeling.  Abstract    Paper (pdf)  R. Subbarao, P. Meer: Heteroscedastic projection based M-estimators.  Abstract    Paper (pdf)  H. Chen, P. Meer: Robust fusion of uncertain information.  Abstract    Paper (pdf)    Paper (ps.gz) B. Georgescu, P. Meer: Point matching under large image deformations and illumination changes.  Abstract    Paper (pdf)  H. Chen, I. Shimshoni, P. Meer: Model based object recognition by robust information fusion.  Abstract    Paper (pdf)    Paper (ps.gz) P. Meer: Robust techniques for computer vision.  Paper (pdf)    Paper (ps.gz) B. Georgescu, I. Shimshoni, P. Meer: Mean shift based clustering in high dimensions: A texture classification example.  Abstract    Paper (pdf)    Paper (ps.gz) H. Chen, P. Meer: Robust regression with projection based M-estimators. Abstract    Paper (pdf)    Paper (ps.gz) H. Chen, P. Meer: Robust fusion of uncertain information. Abstract    Paper (pdf)    Paper (ps.gz)  D. Comaniciu, V. Ramesh, P. Meer: Kernel-based object tracking. Abstract    Paper (pdf)    Paper (ps.gz)      Videos of tracking nonrigid objects.  C. M. Christoudias, B. Georgescu, P. Meer: Synergism in low level vision. Abstract    Paper (pdf)    Paper (ps.gz) H. Chen, P. Meer: Robust computer vision through kernel density estimation. Abstract    Paper (pdf)    Paper (ps.gz) H. Chen, P. Meer, D.E. Tyler: Robust regression for data with multiple structures. Abstract    Paper (pdf)    Paper (ps.gz) P. Meer, B. Georgescu: Edge detection with embedded confidence. Abstract    Paper (pdf)    Paper (ps.gz) D. Comaniciu, P. Meer: Mean shift: A robust approach toward feature space analysis. Abstract    Paper (pdf)    Paper (ps.gz)    ERRATA (pdf)      Test Images used in the paper.   D. Comaniciu, V. Ramesh, P. Meer: The variable bandwidth mean shift and data-driven scale selection  Abstract    Paper (pdf)    Paper (ps.gz)  D. Comaniciu, V. Ramesh, P. Meer: Real-time tracking of non-rigid objects using mean shift. BEST PAPER AWARD     2000 IEEE Computer Vision and Pattern Recognition Conference. Abstract   Paper (pdf)   Paper (ps.gz) P. Meer, C.V. Stewart, D.E. Tyler: Robust computer vision: An interdisciplinary challenge. Abstract   Paper (pdf)   Paper (ps.gz) D. Comaniciu, P. Meer: Mean-shift analysis and applications. Abstract   Paper (pdf)   Paper (ps.gz) D. Comaniciu, P. Meer: Distribution free decomposition of multivariate data. Abstract   Paper (pdf)   Paper (ps.gz)  Examples M. Garza-Jinich, P. Meer and V. Medina: Robust retrieval of 3D structures from image stacks. Abstract   Paper (pdf)   Paper (ps.gz) K-M. Lee, P. Meer and R-H. Park: Robust adaptive segmentation of range images. Abstract   Paper (pdf)   Paper (ps.gz) D. Comaniciu, P. Meer: Robust analysis of feature spaces: Color image segmentation. Abstract   Paper (pdf)   Paper (ps.gz)  Examples Related Ph.D Thesis Dorin Comaniciu: Nonparametric robust Methods for Computer Vision. Bogdan Georgescu: Interpretation of the 3D Visual Environment from Uncalibrated Imagese Sequences. Haifeng Chen: Projection based Robust Estimators for Computer Vision. Raghav Subbarao: Robust Statistics Over Riemannian Manifolds for Computer Vision . Oncel Tuzel: Learning on Riemannian Manifolds for Interpretation of Visual Environments. Sushil Mittal: User-Independent Robust Statistics for Computer Vision. Saket Anand: Robust Methods for Multiple Model Discovery in Structured and Unstructured Data.