Here are a few hundred of the most significant computer vision papers, most recent first. (For a more organized view of these papers see Computer Vision Papers By Topic.)

NB: You may find Google Scholar useful when tracking down references; other useful resources include CiteSeer and arXiv.

- Shen09:
A Duality View of Boosting Algorithms
Shen + Li 2009 15: faster, clearer, stronger, fewer training examples needed.

- Alah08:
Reduce, Reuse & Recycle: Efficiently Solving Multilabel MRFs.
Alahari, Kohli + Torr 2008 8p:
*"10-15X speedup"*

- Bick08:
Distributed Kalman Filter via Gaussian Belief Propagation
Bickson, Shental + Dolev 2008 8p

- Bick08b:
Polynomial Linear Programming with Gaussian Belief Propagation
Bickson, Tock, Shental + Dolev 2008 7p

- Bick08c:
A Gaussian Belief Propagation Solver for Large Scale Support Vector Machines
Bickson, Yom-Tov + Dolev 2008 12p

- Bick08d:
Gaussian Belief Propagation Based Multiuser Detection
Bickson, Dolev, Shental, Siegal + Wolf 2008 5p

- Blai08:
Culture Shapes How We Look at Faces
Blais, Jack, Scheepers, Fiset + Caldara 2008 8p: Yet again,
occidentals are more local, orientals more global.

- Brub08:
Isotropic PCA and Affine-Invariant Clustering
Brubaker + Vempala 2008 27p.

- Cand08:
An introduction to compressive sampling.
Candes + Wakin 2008 9p.

- Chan08:
A Computational Study of the Pseudoflow and Push-relabel Algorithms
for the Maximum Flow Problem.
Chandran + Hochbaum 2008 26p:
*"fastest on many general graphs"*(Delo08)

- Chen08:
Linear Time Recognition Algorithms for Topological Invariants in 3D
Chen + Rong 2008 6p.

- Cour08:
Fast Density Codes for Image Data Courrieu 2008 9p:
Parzen window;
*"shape spaces invariant to a wide variety of affine and non-affine transformations"*;*"Thousands of times faster"*

- Delo08:
A Scalable Graph-Cut Algorithm for N-D Grids. Delong 2008 8p.

- Dinh08:
Two-Frames Accurate Motion Segmentation Using Tensor Voting and Graph Cuts
Dinh + Medioni 2008 8p: Ten minutes per image at present.

- Gort08:
The Lumigraph
Gortler, Grzeszczuk, Szeliski + Cohen 2008 10p:
*"does not reply on a geometric representation"*;*"captures a subset of the plenoptic function"*;*"new images can be generated very quickly"*.

- Kala08:
The Five Points Pose Problem : A New and Accurate Solution Adapted to any Geometric Configuration
Kalantari + Jung 2008 15p: Grobnew bases, Rational Univariate Representation.

- Kohl08:
Robust Higher Order Potentials for Enforcing Label Consistency.
Khohli, Ladicky +
Torr 2008 8p.

- Kohl08a:
Graph cuts for minimizing robust higher order potentials.
Khohli, Ladicky +
Torr 2008 9p.

- Kohl08b:
On Partial Optimality in Multilabel MRFs.
Khohli,
Schekhovtsov, Rother,
**Kolmogorov**+ Torr 2008 8p.

- Koki08:
Classification of multiple observations by semi-supervised learning
Kokipoulou + Frossard 2008 10p:
*"outperforms state-of-the-art solutions" "in video-based face recognition"*

- Kumar08:
Learning Layered Motion Segmentations of Video.
Kumar, Torr +
Zisserman 2008 34p: graph cuts; loopy belief propagation.

- Kuma08b:
Efficiently Solving Convex Relaxations for MAP Estimation.
Kumar+ Torr 2008
8p: Building on tree-reweighted message-passing; optimizing
non-submodular functions; guaranteed convergence; second order cone constraints

- Levy08:
Data-Driven Enhancement of Facial Attractiveness
Leyvand, Cohen-Or, Dror + Lischinski 2008 9p: support vector
regression (SVR); K-nearest neighbors search (KNN); bayesian tangent
shape model (BTSM); multilevel free-form deformation (MFFD).

- Llon08:
3D Face Recognition with Sparse Spherical Representations
LLonch, Kokiopoulou, Tosic + Frossard 2008 23p.

- Mair08:
Supervised Dictionary Learning
Mairal, Bach, Ponce, Sapiro + Zisserman 2008 18p.

- Mant08:
Improved Smoothed Analysis of the k-Means Method
Manthey + Roglin 25:
*"by far the most popular clustering tool used in scientific and industrial applications"*.

- Mogh08:
Sparse Regression as a Sparse Eignvalue Problem
Moghaddam, Gruber, Weiss + Avidan, 2008 7p: 1000X speedup; partitioned
matrix inverse techniques; See also
Mogh07

- Mohi08:
A fast approach for overcomplete sparse decomposition based on smoothed
L0 norm
Mohimani, Babaie-Zadeh + Jutten 2008 30p: Sparse Component Analysis,
*"two to three orders of magnitude faster than the state-of-the-art"*

*Pata08: A Nonparametric Approach to 3D Shape Analysis from Digital Camera Images Patarangenaru, Liu + Sugathadasa 2008 24p: Essential and fundamental matrices;*

*Perc08: The Peculiar Phase Structure of Random Graph Bisection Percus, Istrate, Goncalves, Sumi + Boettcher 2008 24p: Asymptotically within a factor of 1 of optimal.*

*Rama08: Exact Inference in Multi-label CRFs with Higher Order Cliques Ramalingam, Kohli, Alahari + Torr 2008 8p: submodular second-order boolean functions*

*Rava08: Unwrap Mosaics: A new representation for video editing Rav-Acha, Kohli, Rother + Fitzgibbon 2008 11p -- see also the tech report 14p*

*Rost08: Camera distortion self-calibration using the plumb-line constraint and minimal Hough entropy Rosten + Loveland 2008 19p*

*Bhat07: Using Photographs to Enhance Videos of a Static Scene Bhat, Zitnick, Snavely, Agarwala, Agrawala, Curless, Cohen + Kang 2007 12p: superresolution, inpainting, multi-view stereo (MVS), image-based rendering (IBR); structure from motion (SFM); markov random field (MRF); loopy belief propagation; builds on Zitn04.*

*Sang08: Message-passing for Maximum Weight Independent Set Sujay Sanghavi, Devavrat Shah, Alan Willsky 2008 11p.*

*Shen08: Gaussian Belief Propagation Solver for Systems of Linear Equations Shental, Siegel, Wolf, Bickson + Dolev 2008 5p*

*Snav08: Finding Paths through the World's Photos Snavely, Garg, Seitz + Szeliski 2008 11p: image-based rendering (IBR); SIFT; RANSAC; thin-plate splines.*

*Sont08: Tightening LP Relaxations for MAP using message passing Sontag, T. Meltzer, A. Globerson, T. Jaakkola and Y. Weiss 2008 8p: dual message passing.*

*Spar08: An image processing analysis of skin textures Sparavigna + Marazzato 2008 9p*

*Szel08: A Comparative Study of Energy Minimization Methods for Markov Random Fields with Smoothness-Based Priors Szeliski &al 2008 13p: Source code here*

*Vasi08: Learning Isometric Separation Maps Vasiloglou, Gray + Anderson 2008 4p*

*Yang08: Context-aware Visual Tracking Yang, Wu + Hua 2008 15p*

*Yu08: Integrated Detection and Tracking for Multiple Moving Objects using Data-Driven MCMC Data Association. Yu + Medioni 2008 8p: real Adaboost; DDMCMC;*

*Zhou08: Multi-Instance Learning by Treating Instances As Non-I.I.D. Samples Zhou, Sun + Li 2008 16p: MIGraph vs MI-kernel; bags as graphs; Image classification*

*Auja07: Harmonic skeleton for realistic character animation Aujay, Hetroy, Lazarus + Depraz 2007 10p*

*Babe07: Experimental Evaluation of Parametric Max-Flow Algorithms Babenko, Derryberry, Goldberg,***Tarjan**+ Zhou 2007 14p:

*Bara07: Compressive sensing. Baraniuk, 2007 9p:***tutorial**

*Bolm07: FacePerf: Benchmarks for Face Recognition Algorithms Bolme, M Strout, JR Beveridge 2007 6p: with source code.*

*Camp07: Automatic 3D Object Segmentation in Multiple Views using Volumetric Graph-Cuts Campbell, Vogiatzis, Herandez + Cipolla 2007 10p:*

*Gupt07: Efficient inference with cardinality-based clique potentials Gupta, Diwan + Sarawagi 2007 8p: Potts approximation to 13/15 of optimal, up from previous 1/2 of optimal;**"orders of magnitude faster than TRW and graph-cut"*.

*Habb07: A Surface-Growing Approach to Multi-View Stereo Reconstruction Habbecke + Kobbelt 2007 8p*

*Hern07: Probabilistic visibility for multi-view stereo C Hernandez, G Vogiatzis, R Cipolla 2007 8p*

*Hirs07: Evaluation of Cost Functions for Stereo Matching Hirschmuller, D Scharstein 2007 8p: Rank and HMI (Hierarchical Mutual Information) win.*

*Hou07: Multiview Pedestrian Detection Based on Vector Boosting Hou, Ai + Lao, 2007 10p: vector boosting; extended histograms of oriented gradients (EHOG); detector pyramid;**"HOG has proved effective in pedestrian detection"*;

*Hua07: A Decentralized Probabilistic Approach to Articulated Body Tracking Hua 2007 13p.*

*Juan07: Capacity Scaling for Graph Cuts in Vision Juan + Boykov, 2007 8p*

*Kim07: A new graph cut-based multiple active contour algorithm without initial contours and seed points Kim + Hong 2007 6p: swap move + split move to estimate model parameters + region count as part of algorithm..*

*Kohl07: P3 & Beyond: Solving Energies with Higher Order Cliques Kohli, Kumar + Torr, 2007 8p: extended in Kohl07c.*

*Kohl07a: Solving energies with higher order cliques Kohli, Kumar + Torr 2007 10p*

*Kohl07b: Dynamic Graph Cuts for Efficient Inference in Markov Random Fields Kohli + Torr, 2007 8p*

*Kohl07c: Minimizing Dynamic and Higher Order Energy Functions using Graph Cuts Kohli 2007 144p:***dissertation**

*Kolm07: Applications of parametric maxflow in computer vision Kolmogorov, Boykov + Rother 2007 8p: min cut; PDE cuts; ES method; maximum a posteriori - Markov random field (MAP-MRF); trust region graph cuts; image cosegmentation; "The maximum flow algorithm for minimizing energy functions of binary variables has become a standard tool in computer vision."*

*Komo07: Fast, approximately optimal solutions for single and dynamic MRFs Komodakis, Tziritas + Paragios 2007 8p: optimizing non-submodular functions; Fast-PD generalizes alpha-expansion;**"can be at least 3-9 times faster than alpha-expansion"*; continuation of Komo05.

*Kuma07: An Analysis of Convex Relaxations for MAP Estimation Kumar, Kolmogorov + Torr 2007 8p: NP-hard MAP estimation via linear vs quadratic vs second-order cone programming.*

*Kuma07b: An Invariant Large Margin Nearest Neighbour Classifier. Kumar, Torr + Zisserman 2007 8p: Semi-Definite Programming; rotation invariance; convex SDP; face recognition.*

*Leib07: Robust Object Detection with Interleaved Categorization and Segmentation Leibe, Leonarids + Schiele 2007 26p: interest points; codebook generation; implicit shape model (ISM); integrating foreground extraction with object categorization; Generalized Hough Transform; Hough Voting; occlusion handling; minimum description length (MDL); reciprocal nearest neighbor (RNN) agglomerative clustering; Mean Shift; "both rigid and articulated objects"; "training sets ... one [to] two orders of magnitude smaller";*

*Lemp07: Global Optimization for Shape Fitting Lempitsky + Boykov 2007 8p: new regularization functional; touch-expand for large graphs*

*Lemp07b: LogCut-Efficient Graph Cut Optimization for Markov Random Fields Lempitsky, Rother + Blake 2007 8p: new approach for multi-label case: Beating alpha-expansion by being log instead of linear in label count;**"order of magnitude speedup"*; QBPO; non-submodularity.

*Meta07: Physics-Based Modeling Of Nonrigid Objects For Vision and Graphics Metaxas, 2007***thesis**

*Mogh07: Fast Pixel/Part Selection with Sparse Eigenvectors Moghaddam, Yair Weiss, Shai Avidan, 2007 8p: 1000X speedup; sparse eigenfaces; gender classification; See also Mogh08*

*Nowo07: Weighted Substructure Mining for Image Analysis Nowozin + Tsuda 2007 8p: graph mining; linear program boosting; item set boosting; bag-of-words model; interest points; frequent item set mining; support vector machine (SVM); Linear time Closed item set Miner (LCM); codebook; gSpan;*

*Orli07: A faster strongly polynomial time algorithm for submodular function minimization Orlin 2007 13p*

*Osad07: Synergistic Face Detection and Pose Estimation with Energy-Based Models Osadchy, Cun + Miller, 2007 19p: convolutional network; energy based models (EBM);*

*Otoo07: Face Recognition Algorithms Surpass Humans Matching Faces over Changes in Illumination O'Toole, Phillips, Jiang, Ayyad, Penard + Abdi 2007 5p.*

*Raj07: Bayesian Parallel Imaging with Edge-Preserving Priors Raj &al 2007 35: MRI imaging*

*Rama07: Using Segmentation to Verify Object Hypotheses Ramanan 2007 8p: combining detection and segmentation; template classifiers as attention mechanisms;*

*Roth07: Optimizing Binary MRFs via Extended Roof Duality Rother, Kolmogorov, Lempitsky + Szummer 2007 15p: dynamic graph cuts;**"400-700 faster on some graphs"*

*Shar07: A Practical Approach for Super-Resolution using Photometric Stereo and Graph Cuts Sharma, MV Joshi 2007 10p: known light source positions.*

*Sinh07: Multi-View Stereo via Graph Cuts on the Dual of an Adaptive Tetrahedral Mesh Sinha, Mordohai + Pollefeys 2007 8p:*

*Sorm07: Watertight Multi-View Reconstruction Based On Volumetric Graph-Cuts Sormann, Zach, Bauer, Karner + Bishof 2007 10p*

*Tapp07: Utilizing Variational Optimization to Learn Markov Random Fields Tappen 2007 8p: Field of Experts*

*Taub07: Review and Preview: Disocclusion by Inpainting for Image-based Rendering Tauber, Li + Drew 2007 17p*

*Thor07: Fully-Dynamic Min-Cut Thorup 2007 7p*

*Tuze07: Human Detection via Classification on Riemannian Manifolds Tuzel, Porkli + Meer, 2007: boosting via LogitBoost; covariance matrices as object descriptors;**"The space of d-dimensional nonsingular covariance matrices can be represented as a connected Riemannian manifold"**"superior detection rates over previous approaches"*

*Vogi07: Multi-view Stereo via Volumetric Graph-cuts and Occlusion Robust Photo-Consistency Vogiatzis, Esteban, Torr + Cipolla 2007 15p: continued from Vogi05: voting;*

*Xu07: Object Segmentation Using Graph Cuts Based Active Contours Xu, Bansal + Ahuja 2007 8p.*

*Veks07: Graph Cut Based Optimization for MRFs with Truncated Convex Priors Veksler 2007 8p:***range moves**.

*Weis07: What makes a good model of natural images? Weiss + Freeman 2007 8p: non-Guassian prior potentisl; tractable partition function upper/lower bounds; efficient prior learning algorithms; tractable likelihoods; field of experts (FoE) product of experts (PoE); nonintuitive features are real..*

*Wood07: On New View Synthesis Using Multiview Stereo Woodford &al 2007 10p.*

*Wood07b: Efficient new-view synthesis using pairwise dictionary priors Woodford, ID Reid, AW Fitzgibbon 2007 8p: switching from large cliques to pairwise for speed.*

*Yang07: Feature Selection in Face Recognition: A Sparse Representation Perspective Yang, Wright, Ma, Sastry + Shankar 2007 19p: compressed sensing; global representation;**"choice of features no longer critical"**"naturally avoids under- and over-fitting."*

*Yin07: Tree-based Classifiers for Bilayer Video Segmentation Yin, Criminisi, Winn + Essa, 2007 8p: motons; shape-filters; Random Forests; Tree Cube taxonomy of tree-based classifiers; Conditional Random Field (CRF); min-cut;**"enables us to interpret ... AdaBoost, Decision Trees, Random Forests and Cascaded Boosting as variants"**"Correct segmentations are produced even in the presence of large background motion with nearly stationary foreground"*

*Agar06: Practical Global Optimization for Multiview Geometry Agarwal &al UCSD 2006 14p:**"a practical method for finding the provably globally optimal solution to ... multiview triangulation, camera resectioning and homography estimation"*;*"relies on recent developments in fractional programming and the theory of convex underestimators"*;*"open source MATLAB toolbox"*;*"Projective geometry is one of the success stories of computer vision"*;*"work horse subroutines ... matchmove in filmmaking to ... image mosaicing"**"branch and bound style method"**"worst case complexity is exponential"*but almost linear in practice.

*Agar06b: A Local Basis Representation for Estimating Human Pose from Cluttered Images Agarwal + Triggs 2006 10p: bottom up; SIFT-like histograms; gradient orientation histograms; non-negative matrix factorization (NMF); direct regression; does not require prior segmentation.*

*Agar06c: Recovering 3D Human Pose from Monocular Images Agarwal + Triggs 2006 15p: errata; silhouettes; Relevance Vector Machines; 54-parameter model;**"a factor of three better"*

*Ahon06: Face Description with Local Binary Patterns: Application to Face Recognition Ahonen, Hadid + Pietikainen 2006 15p.*

*Appl06: Globally minimal surfaces by continuous maximal flows Appleton + Talbot 2006 26p.*

*Babe06: Experimental Evaluation of a Parametric Flow Algorithm Babenko + Goldberg 2006 12p: This appears to be an early draft of Babe07*

*Boyk06: An Integral Solution to Surface Evolution PDEs via Geo-Cuts Boykov, Kolmogorov, Cremers + Delong 2006 14p: Improves on level-set methods; handles topological changes.*

*Boyk06b: From Photohulls to Photoflux Optimization Boykov + Lempitsky 2006 10p:**"unifies space-carving and deformable models"*: builds on Kutu00.

*Boyk06c: Graph Cuts and Efficient NDImage Segmentation Boykov + Funka-Lea 2006 23p.*

*Bray06: PoseCut: Simultaneous Segmentation and 3D Pose Estimation of Humans using Dynamic Graph-Cuts. Bray, Kohli + Torr 2006 14p.*

*Cand06: Compressive sampling Candes 2006 20p.*

*Cran06: Weakly Supervised Learning of Part-Based Spatial Models for Visual Object Recognition Crandall + Huttenlocher, Cornell 2006, 14p:**"our weakly supervised technique produces better results than thes previous highly supervised methods"*; Markov random field; expectation maximization (EM); maximum a posteriori (MAP); patch models;

*Crem06: Statistical Priors for Efficient Combinatorial Optimization via Graph Cuts Cremer + Grady 2006 12p:**"hundreds of statistically learned constraints per pixel"*.

*Crim06: Bilayer Segmentation of Live Video Criminisi &al 2006 8p: optical flow analysis not needed; Hidden Markov Model; Markov Random Field; graph cut;**"accuracy comparable to state of the art stereo segmentation"*

*Dala06: Human Detection Using Oriented Histograms of Flow and Appearance Dalal, Tiggs + Schmid 2006 INRIA 14p*

*Ever06: The PASCAL Visual Object Classes Challenge 2006 (VOC2006) Results Everingham + Zisserman 2006 57p: See also Ever05*

*Gal06: Salient Geometric Features for Partial Shape Matching and Similarity Gal + Cohen-Or 2006 23p: Triangulation-independent surface matching; geometric hashing; voting scheme;*

*Grab06: Real-Time Tracking via On-line Boosting Grabner, Grabner + Bischof 2006 10p: "AdaBoost feature selection" "adapt the classifier while tracking" "depending on background ... selects the most discriminating features" "runs in real-time"*

*Grab06b: Online Boosting and Vision Grabner + Bischof, 2006, 8p -- focussing on discrete AdaBoost.*

*Grau06: Approximate Correspondences in High Dimensions Grauman + Darrel, MIT, 2006, 12p: Pyramid embedding; non-uniform bins; linear time; Mercer kernel; partial matching; histograms*

*Grub06: Incorporating non-motion cues into 3D motion segmentation Gruber + Weiss 2006 14p: measurement matrix factorization; constrained factor analysis; EM; Potts model; graph cuts*

*He06: Learning and Incorporating Top-Down Cues in Image Segmentation He, Zemel + Ray 2006 14p: conditional random field; object segmentation; superpixels: Follow-up of He04.*

*Hega06: Patch-based Object Recognition Using Discriminatively Trained Gaussian Mixtures Hegarath, Deselaers + Ney 2006 10p: generative model; PCA; wavelet;**"recently ... patch-based models ... enormous amount of interest"*:

*Heng06: Rapid Interactive Modelling from Video with Graph Cuts Hengel, Dick, Thormaehlen, Ward + Torr 2006 4p: dynamic graph cuts*

*Hirs06: Stereo Vision in Structured Environments by Consistent Semi-Global Matching Hirschmuller 2006 8p: Continuation of Hirs05*

*Horn06: Hierarchical Volumetric Multi-view Stereo Reconstruction of Manifold Surfaces based on Dual Graph Embedding Hornung + Kobbelt 2006 8p*

*Hua06: Probabilistic Variational Methods for Vision based Complex Motion Analysis" Hua 2006 176p***dissertation**

*Huss06: Real-Time Human Detection, Tracking and Verification in Uncontrolled Camera Motion Environments Hussein &al 2006 7p*

*Juan06: Active Graph Cuts Juan + Boykov 2006 7p: s-t cuts;**"can effectively use a good approximate solution (initial cut)"*;*"faster than the state-of-the-art max-flow methods"*

*Kohl06: Measuring Uncertainty in Graph Cut Solutions Kohli + Torr 2006 14p*

*Kolm06: Comparison of energy minimization algorithms for highly connected graphs Kolmogorov + Rother 2006 16p:**"For our problem graph cut outperforms both TRW and BP considerably"*

*Kolm06b: Minimizing Nonsubmodular Functions with Graph Cuts -- A Review Kolmogorov + Rother 2006 15p*

*Kuma06: Solving Markov Random Fields using Second Order Cone Programming Relaxations Kumar, PHS Torr, A Zisserman 2006 8p: 0-1 quadradic programming;**"significantly outpeforms semidefinite relaxations"*;

*Kuri06: Unsupervised segmentation of words into morphemes -- Challenge 2005 An Introduction and Evaluation Report Kurimo, Creutz + Varjokallio 2006 56p*

*Lafo06: Diffusion Maps and Coarse-Graining: A Unified Framework for Dimensionality Reduction, Graph Partitioning and Data Set Parameterization Lafon + Lee 2006 26p: clustering; nonlinear dimensionality reduction; markov random walks; reorganizing graphs.*

*Lapt06: Improvements of Object Detection Using Boosted Histograms Laptev 2006 10p INRIA*

: AdaBoost; Weighted Fisher Linear Discriminant (WFDL): "outperforms*all*methods reported"*Le06: Ent-Boost: Boosting using entropy measures for robust object detection Le + Satoh 2006 8p: Kullback-Leibler divergence;**"Class entropy information is used to automatically estimate the optimal number of bins"*;*"good performance"**"compact storage space"*

*Lemp06: Oriented visibility for multiview reconstruction Lempitsky, Boykov + Ivanov 2006 12p*

*Levi06: Learning to combine bottom-up and top-down segmentation Levin + Weiss 2006 15p: CRF; TRW; cows; octopi*

*Liu06: Capitalize on Dimensionality Increasing Techniques for Improving Face Recognition Grand Challenge Performance Liu 2006 13p: outperformed humans in Otoo07*

*Mart06: View Synthesis for Multiview Video Compression Martinian et al 2006 5p*

*Miko06: Multiple Object Class Detection with a Generative Model Mikolajczyk, Leibe + Schiele 2006 8p: codebook; "excellent performance on several object categories over a wide range of scale, in-plane rotations, background clutter and partial occlusions"; builds on Lowe03.*

*Mitc06: The Discipline of Machine Learning Mitchell, CMU, 2006, 12p*

*Mude06: Super resolution using graphcut Mudenagudi, Singla, Kalra, + Banerjee 2006 6p*

*Nowa06: Sampling Strategies for Bag-of-Features Image Classification Nowak, Jure + Tiggs 2006, 124*

*Pari06: A Surface Reconstruction Method Using Global Graph Cut Optimization Paris 2006, 36p: multi-view, 1/10 pixel triangle mesh reconstruction, occusion handling.*

*Perl06: Face Recognition using Principal Component Analysis and Log-Gabor Filters Perlibakas 2006, 23p: "simple implementation, training and very high recognition rate".*

*Perl06b: Recognition of expression variant faces using masked log-Gabor features and Principal Component Analysis Perlibakas 2006, 20p: "log-Gabor filters of multiple orientations and scales" "96.6-98.9% recognition rate".*

*Phil06: Preliminary Face Recognition Grand Challenge Results Phillips, Flynn, Scruggs, Bowyer + Worek 2006 7p.*

*Rabi06: Model Order Selection and Cue Combination for Image Segmentation Rabinovich, Lange, Buhmann + Belongie 2006, 8p*

*Riha06: OBJCUT for face detection Rihan, Kohli + Torr 2006 10p: face segmentation; real-time; graph cuts; coupled face detection and segmentation; match uncertainty measure; 2nd-gen CRF + shape prior;**"extremely computationally intensive"*(Kohl07c)

*Raj06: MRFs for MRIs: Bayesian reconstruction of MR images via graph cuts Raj, Singh + Zabihh 2007 10p*

*Ross06: Model-free statistical detection and tracking of moving objects Ross 2006 4p: particle filter; occlusion handling*

*Roth06: Cosegmentation of Image Pairs by Histogram Matching -- Incorporating a Global Constraint into MRFs Rother, Kolmogorov, Minka + Blake 2006 14p: generative model; iterated graph cuts;**"trust region graph cuts"*;*"potential to improve ... image retrieval, video tracking and segmentation, interactive image editing..."*;*"the power of the framework lies in its generality"*

*Russ06: Using multiple segmentations to discover objects and their extent in image collections Russell, Efros, Sivic, Freeman + Zisserman 2006 8p: statistical text analysis;**"grouping visual words"*

*Scar06: A Flexible Technique for Accurate Omnidirectional Camera Calibration and Structure from Motion Scaramuzza &al 2006 8p*

*Schl06: Transforming an arbitrary minsum problem into a binary one Schlesinger + Flach 2006 16p*

*Schm06: Kernel Particle Filter for Real-Time 3D Body Tracking in Monocular Color Images Schmidt + Fritsch 2006 6p*

*Scho06: Near real-time motion segmentation using graph cuts Schoenemann + Cremers 2006 10p: Gaussian-distributed region velocties; linearized flow errors for speed;*

*Schr06: Single-Histogram Class Models for Image Segmentation Schroff, Criminisi + Zisserman 2006, 13p: cows; textons; Kull-Leibler divergence beats Euclidean distance; K-means clustering;*

*Seit06: A Comparison and Evaluation of Multi-View Stereo Reconstruction Algorithms Seitz &al 2006 8p*

*Shot06: Textonboost: Joint appearance, shape and context modeling for multi-class object recognition and segmentation Shotton, Winn, Rother + Criminisi 2006 14:**"joinly model shape and texture"*; CRF; cows

*Skre06: Texture classification using sparse frame-based representations Skretting + Husoy 2006 6p:*

*Sinh06: Face Recognition by Humans: Nineteen Results All Computer Vision Researchers Should Know About Sinha &al 2006 15p*

*Stic06: Global Depth from Epipolar Volumes--A General Framework for Reconstructing Non-Lambertian Surfaces Stich, Tevs + Magnor 2006 8p: Epipolar Volumes;**"reconstruct highly specular surfaces"*; occlusion handling; Plenoptic function; 3X speedup via GPU; alpha-expansion tweak; POVray

*Sun06: Using Strong Shape Priors for Stereo Sun, Kohli, Bray + Torr 2006 12p:***best paper award**

*Sun06b: Background cut Sun, W Zhang, X Tang, HY Shum 2006 14p: background contrast attenuation; min-cut; video teleconferencing.*

*Szel06: A Comparative Study of Energy Minimization Methods for Markov Random Fields Szeliski &al 2006 17p: graph cuts, loopy belief propagation (LBP), tree-reweighted message passing, iterated conditional modes (ICM); "on photomontage expansion moves perform best"; "there never seems to be any reason to use swap moves instead of expansion moves"; "on the inpainting benchmark TRW-S is the winner"; "LBP performed surprisingly poorly"; "TRW-S, which has not been widely used in vision, gave consistently strong results"*

*Szel06b: Image alignment and stitching: a tutorial Szeliski 2006 89p*

*Tess06: A real-time optical head tracker based on 3D prediction and correction Tessens, Kehl, Pizurica, Gool + Philips 2006 4p: 28Hz, Levenberg-Marquardt*

*Ultr06: 3D People Tracking with Gaussian Process Dynamical Models Utrasun, Fleet + Fua 2006: Gaussian Process Dynamical Models (GPDM), pose probability estimation. "It provides a continuous density function over poses and motions that is generally non-Gaussian and multimodal."*

*Wang06: The cartoon animation filter Wang, Drucker, Agrawala + Cohen 2006 5p:**"simplicity and generality"*; generates*"anticipation, follow-through, exaggeration and squash-and-stretch"*

*Will06: A Switched Gaussian Process for Estimating Disparity and Segmentation in Binocular Stereo Williams 2006.*

*Wu06: Tracking of Multiple, Partially Occluded Humans based on Static Body Part Detection Wu + Nevatia 2006: Continued from Wu05.*

*Wu06b: Dense Photometric Stereo: A Markov Random Field Approach Wu, Tang, Tang + Wong 2006 17p.*

*Yang06: Stereo Matching with Color-Weighted Correlation, Hierachical Belief Propagation and Occlusion Handling Yang, Wang, Yang, Stewenius + Nister 2006.*

*Yuil06: Vision as Bayesian inference: analysis by synthesis? Yuille + Kersten 2006.*

*Zhan06: Local Features and Kernels for Classification of Texture and Object Categories: A Comprehensive Study Zhang, Lazebnik + Schmid 2006 26p: Support Vector Machine (SVM); Earth Mover's Distance (EMD); histograms;**"exceeds best reported results"*

*Zhu06: Fast Human Detection Using a Cascade of Histograms of Oriented Gradients Zhu &t al UCSB 2006 8p*

*Aran05: Automatic Face Recognition for Film Character Retrieval in Feature-Length Films Arandjelovic + Zisserman 2005 8p*

*Aran05b: Face recognition with image sets using manifold density divergence Arandjelovic, Shakhnarovich, Cipolla + Darrel 2005 8p: Gaussian Mixture Models, low-dimensional non-linear manifolds; Kullback-Leibler divergence; Expectation Maximization (EM); Viola-Jones;**"good results even when test illumination not present in training set"*

*Barb05: Cluster Sampling and Its Applications in Image Analysis Barbu + Zhu 2005 44p: Markov Chain Monte Carlo (MCMC) methods: "Extends Swendsen-Wang to general Bayesian inference on graphs"; image segmentation; motion analysis; see also Barb05b.*

*Barb05b: Generalizing Swendsen-Wang to Sampling Arbitrary Posterior Probabilities Barbu + Zhu 2005 35p: arbitrary energy functions; infer generative models; better results than graph cuts or belief propagation; image segmentation; stereo vision. Continuation of Barb05.*

*Berg05: Shape matching and object recognition using low distortion correspondence Berg + Malik 2005 35.*

*Boim05: Detecting Irregularities in Images and in Video Boiman + Irani 2005 8p: Maximum A Posteriori (MAP) probability; Belief Propagation; faster than "constellation" approach; coarse-to-fine;*

*Bowy05: A survey of approaches and challenges in 3D and multi-modal 3D + 2D face recognition Bowyer, Chang + Flynn, 2005, 15p*

*Bron05: Three-Dimensional Face Recognition Bronstein, Bronstein + Kimmel 2005 26: expression-invariant representations; can distinguish identical twins; avoids surface reconstruction; Riemannian surfaces; metric tensors; 3DFACE; Fast Marching Method (FMM).*

*Chen05: Computer Vision Workload Analysis: Case Study of Video Surveillance Systems*

Chen et al, Intel Technology Journal V09:02 , 2005 12p

This is an overview of both the**blobtracker**algorithms and code, and also computer vision more generally, with many good references to both tutorial and research papers.

*Cheu05: Video epitomes Cheung, Frey + Jojic UToronto 2005 8p: Dirac functions; video interpolation; super-resolution; missing data reconstruction*

*Culo05: Gene prediction with conditional random fields Culotta, Kulp + McCallum 2005 14p*

*Dala05: Histograms of Oriented Gradients for Human Detection Dalal + Triggs 2005 8p.: Histograms of Oriented Gradiants (HOG); Support Vector Machine (SVM);*

*Debo05: Scalable partitioning and exploration of chemical spaces using geometric hashing Dutta &al 2005 13p: fast approximate nearest-neighbor search via Locality Sensitive Hashing (LSH -- 100X faster than k-nearest-neighbor, 94% accurate)*

*Deng05: A symmetric patch-based correspondence model for occlusion handling Deng, Yang, Lin + Tang 2005 7p.*

*Drou05: Geo-consistency for Wide Multi-Camera Stero Drouin, Trudeau + Roy 2005 8p.*

*Ever05: The 2005 PASCAL Visual Object Classes Challenge Everingham &al 2005 65p: See results in Ever06*

*Felz05: Pictorial Structures for Object Recognition Felzenszwalb + Huttenlocher 2005 42p: See also Kuma05, Kuma04, Felz00.*

*Free05: Energy minimization via graph cuts: Settling what is possible D Freedman, P Drineas 2005 8p: higher order energy functions.*

*Grad05: Multilabel Random Walker Image Segmentation Using Prior Models Grady 2005 8p:**"deep connection with graph cuts"*.

*Hamp05: Smart Video Surveillance A Hampapur, L Brown, J Connell, A Ekin, N Haas 14p 2005 (magazine survey article): background subtraction, 2D blob tracking, 3D blob tracking, head pose detection, face detection.*

*He05: Face Recognition Using Laplacianfaces He, Yan, Hu + Niyogi 2005 34p:**more discriminating than PCA, less sensitive to outliers"*.

*Hirs05: Accurate and Efficient Stereo Processing by Semi-Global Matching and Mutual Information Hirschmuller 2005 8p: the seminal Semi-Global Matching paper; occlusion handling; 100MPixel aerial photos. Continued in Hirs06*

*Hoie05: Automatic Photo Pop-up Hoiem. Efros. Hebert 2005 8p: single view reconstruction, superpixels.*

*Hoie05b: Geometric Context from a Single Image Hoiem. Efros. Hebert 2005 8p: single view reconstruction, superpixels.*

*Huan05: Vector Boosting for Rotation Invariant Multi-View Face Detection Huang, Ai, Li + Lao 2005 8p: Likely a continuation of Huan04; WFS tree, "one of the most efficient methods"(Hou07)*

*Raj05: A Graph Cut Algorithm for Generalized Image Deconvolution Raj, Singh + Zabih 2007 10p: MRI reconstruction*

*Kehl05: Full Body Tracking from Multiple Views Using Stochastic Sampling Kehl, Bray + Gool 2005 8p*

*Kehl05b: Markerless Full Body Tracking by Integrating Multiple Cues Kehl, Bray + Gool 2005 8p*

*Kesh05: A Simpler, Intuitive Approach to Morpheme Induction Keshava + Pitler 2005 5p: Performed best on English according to Kuri06*

*Kohl05: Efficiently Solving Dynamic Markov Random Fields Using Graph Cuts Kohli + Torr, 2005 8p*

*Kolm05: Probabilistic fusion of stereo with color and contrast for bi-layer segmentation Kolmogorov, Criminisi, Blake + Rother, 2005 18p: Layered Dynamic Programming (LDP); Layered Dynamic Cut (LGC)*

*Kolm05b: Convergent Tree-Reweighted Message Passing for Energy Minimization Kolmogorov, 2005 24p: stereo; min-marginals; Markov Random Fields; Improves on Wain03 via better convergence properties. Can be implemented in half the space of the usual Belief Propagation, which can be seen as a special case.(Szel06); sequential tree-reweighted message passing; non-decreasing bound guarantee;**"outperforms both ordinary belief propagation and tree reweighted algorithm"*

*Kolm05c: What Metrics Can Be Approximated by Geo-Cuts, or Global Optimization of Length/Area and Flux Kolmogorov, 2005 8p*

*Kolm05d: On the optimality of tree-reweighted max-product message passing Kolmogorov + Wainwright, 2005 8p*

*Kolm05e: Bi-layer segmentation of binocular stereo video Kolmogorov, Criminisi, Blake, Cross + Rother 2005 8p: video is here; dynamic programming; ternary graph cut.*

*Kolm05f: Primal-dual Algorithm for Convex Markov Random Fields Kolmogoriv 2005 17p:**"for the panoramic stitching problem our method outperforms other techniques"*.

*Kolm05g: Graph Cut Algorithms for Binocular Stereo with Occlusions Kolmogoriv + Zabi 2005 17p.*

*Komo05: A new framework for approximate labeling via graph cuts Komodakis + Tziritas 2005 8p: generalizes alpha-expansion; alternate interpretation of alpha-expansion as special case of primal-dual: See also Komo07.*

*Komo05b: Approximate labeling via the primal-dual schema Komodakis + Tziritas 2005 29p.*

*Korh05: Particle Filtering in High Clutter Environments Korhonen et al, 2005 4p*

*Kuma05: OBJCUT Kumar, Torr + Zisserman 2005 29p: cows; combining top-down + bottom-up cues; pictorial structures; graph cuts; Potts model; See also See also Felz05, Kuma04, Felz00*

*Lan05: Beyond Trees: Common-Factor Models for 2D Human Pose Recovery Lan + Huttenlocher 2005 8p: latent variables; limb coordination; factor analysis; Viterbi algorithm*

*Lan05b: Efficient Belief Propagation with Learned Higher-Order Markov Random Fields Lan, Roth, Huttenlocher + Black 2005 14p*

*Lapt05: Periodic Motion Detection and Segmentation via Approximate Sequence Alignment Laptev &al 2005 UCSD 8p: fundamental matrix; RANSAC;*

*Li05: Video object cut and paste Li, Sun + Shum 2005 6p: 3D graph-cut of video volumes; watershed regions; bidirectional feature tracking; coherent matting; alpha mattes*

*Mats05: Practical Efficiency of Maximum Flow Algorithms Using MA Ordering and Preflows Matsuoka + Fujishige 2005 8p: A speed-up of Fuji03.*

*Melt05: Globally optimal solutions for energy minimization in stereo vision using reweighted belief propagation Meltzer, Yanover + Weiss 2005 8p: "global optimum can be found in about 30 minutes" -- but the resulting solution still has serious problems.*

*Miko05: A Performance Evaluation of Local Descriptors Mikolajczyk + Schmid 2005 16p: SIFT wins; improved SIFT.*

*Mori05: Recovering 3d Human Body Configurations Using Shape Contexts Mori + Malik 2005 24p: Continued from Mori04*

*Neub05: 3D Texture Reconstruction from Extensive BTF Data Neubeck, A Zalesny, LV Gool 2005 6p: Exemplar based; handling 40GB datasets;**"viewpoint robustness of reflectance features"*

*Oze05: Online Bagging and Boosting Oza 2005 6p: Converting batch bagging and boosting algorithms to online use.*

*Page05: An Automatic 3D Texturing Framework Paget 2005 6p*

*Pape05: Highly Accurate Optic Flow Computation with Theoretically Justified Warping Papenberg 2005 18p*

*Phil05: Overview of the Face Recognition Grand Challenge Phillips, Flynn, Scruggs, Bowyer, Chang, Hoffman, Marques, Min + Worek 2005 8p.*

*Plue05: STACS: New Active Contour Scheme for Cardiac MR Image Segmentation Pluempitiwriyawij, Moura, Wu + Hog 2005 11p: random fields; energy functions*

*Rico05: Large scale stereo reconstruction by a minumum s-t cut formulation Rico 2005 112p***master's thesis**

*Rive05: Entropy controlled Gauss-Markov random measure field models for early vision Rivera, Ocegueda + Marroquin 2005 8p: efficient segmentation-restoration; faster.*

*Rose05: A System for Marker-Less Human Motion Estimation Rosenhahn &al 2005 8p*

*Roth05: Digital Tapestry Rother, Kumar, Kolmogorov + Blake 2005 8p: extending expansion move with non-metric hard and soft constraints.*

*Roth05b: A framework for learning image priors Roth + Black 2005 8p: Fields of experts: image denoising; image inpainting; Products-of-Experts (PoE); Field-of-Experts (FoE); sparse coding; translation-invariant priors; higher-order Markov Random Fields;**"extended pixel neighborhoods"**"all parameters learned from training data"*; gradient ascent

*Sinh05: Multi-view Reconstruction using Photo-consistency and Exact Silhouette Constraints: A Maximum-Flow Formulation Sinha + Polleyfeys 2005 8p*

*Smin05: Conditional models for contextual human motion recognition Sminchisescu et al 2005: Conditional Random Fields (CRF) and maximum entropy Markov models (MEMM) vs hidden Markov model (HMM).*

*Sanf05: An approach to visual motion analysis Sanfeliu + Villanueva 2005 14p: CONDENSATION; iTrack; Cue integration; particle filtering;*

*Sere05: Object Recognition with Features Inspired by Visual Cortex Serre, Wolfe + Poggio 2005 7p MIT: gentle AdaBoost; linear SVM; online source code;**"able to learn from very few examples"*;*"outperforms several state-of-the-art systems"**"performs significantly better than all systems we have compared it with"*

*Shei05: Bayesian object detection in dynamic scenes Sheikh, Shah 2005 6p: pixel correlations, temporal persistence, foreground modelling.e*

*Stew05: A Minimal Solution for Relative Pose with Unknown Focal Length Stewenius &al 2005 6p: one scene two cameras; eigen-decomposition of 15x15 matrix.*

*Tian05: Robust and Efficient Foreground Analysis for Real-time Video Surveillance Tian, Lu + Hampapur 6p 2005: Mixture of Gaussians (MoG); quick lighting changes*

*Torr05: Representational oriented component analysis (ROCA) for face recognition with one sample image per Training Class Torre, Gross, Baker + Kumar 8p 2005: subspace methods (SM);**"outperforms PCA, OCA and some commercial systems"**"avoids overfitting"*

*Tu05: Image Parsing: Unifying Segmentation, Detection, and Recognition Tu, Chen, Yuille + Zhu 2005 28p: generating parsing graph via reversable Markov Chain jumps; integrating generative and descriminative methods; texture + shading; human faces; text; Metropolis-Hastings; PCA; Adaboost; DDMCMC.*

*Urta05: Priors for people tracking from small training sets Urtasun, Fleet, Hertzmann + Fua 2005 8p: golfing, walking;**impressive pose inference from monocular data"*(Kohl07c)

*Varm05: A Statistical Approach to Texture Classification from Single Images Varma + Zisserman 2005 34p: rotationally invariant filters; low dimensional filter space; textons; histograms;**"superior performance"*

*Vogi05: Multi-view Stereo via Volumetric Graph-cuts Vogiatzis, Torr + Cipolla 8p 2005: continued in Vogi07*

*Wern05: A Linear Programming Approach to Max-Sum Problem: A Review Werner 2007 46p.*

*Yaun05: Volume cutout Yuan, Zhang, Nguyen + Chen 2005 9p: two-pass graph cuts algorithm;*

*Wain05: MAP Estimation Via Agreement on Trees: Message-Passing and Linear Programming Wainwright, Jaakkola + Willsky 2005 25p*

*Wang05: Interactive video cutout Wang, Bhat, Colburn, Agrawala + Cohen 2005 10p: oliphants; hierarchical mean-shift minimization of min-cut nodes; local min-cut cost functions; 3D volume matting.*

*Wern05: Image Segmenation Using Minimum st Cut Werner 18p 2005(?):***slides**

*Wolf05: Robust Boosting for Learning from Few Examples Wolf + Martin 2005 6p: gentleBoost; noise injection; feature knockout;**"enhancing our dataset with corrupted copies of our original data"*; boosting more likely to overfit than SVM;*"the major disadvantage of SVM is that it uses at runtime when classifying a new example X all the measurements (features) of x ... for object detection problems in vision ... we often need to search the whole image in several scales over thousands of locations".*;

*Wu05: Detection of Multiple, Partially Occluded Humans in a Single Image by Bayesian Combination of Edgelet Part Detectors Wu + Nevatia 2005 8p: Seminal edgelets paper, continued in Wu06: Detecting walking and standing figures in static images; Maximum A Posteriori (MAP) estimation; Real AdaBoost; histogram weak classifiers; nesting structured detector (vs cascade)*

*Xiao05: Motion Layer Extraction in the Presence of Occlusion using Graph Cuts Xiao + Shah 2005 35p*

*Yu05: Shadow graphs and 3 D texture reconstruction Yu + Chang 2005 42p: Shape from shading; Shape from shadow.*

*Zhan05: Simultaneous Parametric Maximum Flow Algorithm with Vertex Balancing Zhang, Ward + Feng 2005 8p*

*Agar04: Interactive digital photomontage Agarwala &al 2004 9p: Interactive Digital Photomontage: graph cut; gradient-domain fusion; panoramic stitching; extended depth of field; time-lapse mosaics; relighting.*

*Agar04b: 3D human pose from silhouettes by relevance vector regression Agarwal + Triggs 2004 7p.*

*Ahon04: Face recognition with local binary patterns Ahonen, Hadid + Pietikainen 2004 13p: Nearest neighbor via Chi-square dissimilarity measure.*

*Alle04: Object Tracking Using CamShift Algorithm and Multiple Quantized Feature Spaces Allen, Xu + Jin 2004 5p: Moving beyond the OpenCV camshift-based facetracker. Continuously Adaptive Mean Shift (CamShift); HSV color histogram; histogram back-projection;*

*Barb04: Multi-grid and multi-level Swendsen-Wang cuts for hierarchic graph partition Barbu + Zhu 2005 8p.*

*Blak04: Interactive image segmentation using an adapative GMMRF model Blake, Rother, Brown, Perez + Torr 2004 14p*

*Boyk04: Optimal Object Extraction via Constrained Graph-Cuts Boykov + Funka-Lea 2004 ?p: (I can't find a copy of this online.)*

*Cheu04: Robust techniques for background subtraction in urban traffic video Cheung + Kamath 2004 12p: Surveys a variety of background subtraction algorithms, claims simple adaptive median filtering should not be underestimated.*

*Corr04: 3D Human Posture Estimation Using Geodesic Distance Maps Correa, Marques, Marichal + Macq 2004 4p: geodesic distance to find extremeties, morphological skeleton.*

- Crem04:
Kernel Density Estimation and Intrinsic Alignment for Knowledge-driven
Segmentation: Teaching Level Sets to Walk Cremers, Osher + Soatto
2004 8p; see also Crem06.

- Cris04: A
Multi-Stage Approach to Facial Feature Detection: Cristinacce,
Coots + Scott 2004 10p: Pairwise Reinforcement of Feature Responses
(PRFR); boosted cascade classifier; Active Appearance Model (AAM);
nonparametric pairwise histograms;

- Cucc04: Track-based and object-based occlusion for people tracking refinement in indoor surveillance: Cucchiara, Grana + Tardini 2004 7p

- Fei04: Joint Bayes Filter: A
Hybrid Tracker for Non-Rigid Hand Motion Recognition Fei & Reid
2004 12p: Hidden Markov Model; Particle filter; generative model;
occlusion; Viterbi algorithm.

- Feif04:
Learning Generative Visual Models from Few Training Examples: An
Incremental Bayesian Approach Tested on 101 Object Categories
Fei-Fei, Fergus + Perona 2004 9p:
*"learning categories from just a few training examples"*;*"quick"*;*"constellation of features"*;*"outperforms maximum likelihood"*;

- Felz04: Efficient Belief
Propagation for Early Vision Felzenszwalb + Huttenlocker, 2004,
8p, Cornell: detailed description of
Loopy Belief Propagation; Markov Random Fields; optic flow; stereo
matching; image restoration;
*"Taken together, these techniques speed up the standard algorithm by several orders of magnitude."*;

- Felz04b: Efficient
Graph-Based Image Segmentation Felzenszwalb 2004 26p MIT:
Low-level image segmentation; Emphasis on linear running time and practical application.

- Felz04c:
Distance Transforms of Sampled Functions
Felzenszwalb + Huttenlocher 2004 15p: Kolm07c
uses this.

- Grad04: Isoperimetric Graph Partitioning for Data Clustering and Image Segmentation Grady + Schwartz 2004 30p: 10X faster than Ncut.

- Grad04b: Multi-Label Image Segmentation for Medical Applications Based on Graph-Theoretic Electrical Potentials Grady + Funka-Lea 2004 13p: graph random walks

- Grad04c: Faster
Graph-Theoretic Image Processing via Small-World and Quadtree
Topologies Grady + Schwartz 2004 6p: isoperimetric algorithm;
image segmentation;
*"large speedup with minimum perturbation of the solution"*

- Hamz04:
From Fields to Trees
Hamze + Freitas 2004 8p: Rao-Blackwellized Markov Chain Monte Carlo
Methods, a more principled alternative to Swendsen-Wang &tc; works
*"even when Loopy Belief Propagation fails to converge"*.

- He04:
Multiscale Conditional Random Fields for Image Labeling
He, Zemel + Carreira-Perpina 2004 8p: see also He06".

- Hong04:
Segment-based stereo matching using graph cuts
Hong + Chen 2004 8p: moving from pixel domain to segment domain.

- How04:
Better Foreground Segmentation Through Graph Cuts
Howe + Deschamps 2004 10p: with MATLAB source code.

- Huan04: Boosting
Nested Cascade Detector for Multi-View Face Detection Huang, Ai, Wu
+ Lao 2004 4p: Continued in Huan05

- Huan04b:
Omni-Directional Face Detection Based on Real AdaBoost Huang, Wu,
Ai + Lao 2004 4p: Continued in Huan05; Haar-type
features; flip and rotate detectors;

- Kehl04:
Real-time Pointing Gesture Recognition for an Immersive Environment
Kehl + Gool 2004 6p

- Kirs04:
A discrete global minimization algorithm for continuous variational problems
Kirsanov + Gortler
2004 21p

- Kolm04:
What Energy Functions Can Be Minimized via Graph Cuts?
Kolmogorov + Zabih 2004 30p

- Kuma04:
Extending pictorial structures for object recognition
Kumar, Torr + Zisserman 2004 10p: cows; generative models; exemplar-based models; tree cascade
of classifiers; coarse-to-fine; chamfer distance; loopy belief
propagation; see also Felz05, Kuma05, Felz00

- Laze04:
A Sparse Texture Representation Using Local Affine Regions
Lazebnik, Schmid + Ponce 2004 35p: textons; spin image; RIFT descriptor; Earth
Mover's Distance (EMD); k-means; histograms.

- Levi04:
Learning Object Detection from a Small Number of Examples: the
Importance of Good Features Levi + Weiss 2004 8p: Lapt06 improves upon this.

Frontal and profile faces; local edge orientation histograms; AdaBoost; Sobel masks; symmetry detection;*"for profile view faces ... seems to outperform the state of the art in realtime systems"*. - Mori04:
Recovering human body configurations: combining segmentation and recognition
Mori, Ren, Efros + Malik 2004 8p: See also Mori05.

- Peng04:
Accurate information extraction from research papers using conditional random fields
Peng + McCallum 2004 8.

- Mitt04: Motion-Based Background
Subtraction Using Adaptive Kernel Density Estimation Mittal +
Paragios 2004 8p: Adding optical flow to the background model to
better handle dynamic backgrounds.

- Opel04: Generic
Object Recognition with Boosting Opelt, Fussenegger, Pinz + Auer
2004 22p: Harris-Laplace detector; affine invariant interest point
detector; difference-of-Gaussians (DoG) keypoint detector; SIFT;
Homomorphic Filtering; Fast Fourier Transform (FFT);
*"allows segmentation of non-connected regions"*;*"dramatically improves the stability of a classifier"*; Introduces Similarity-Measure-Segmentation;*"equal or better than state-of-the-art segmentation techniques"*;*"we aim handling complex objects in highly cluttered images"**"outperforms all comparable solutions"*.

- Pari04:
Extraction of Three-dimensional Information from Images
Paris 2004 232p:
**dissertation**: graph cuts; face relighting; capture of hair geometry.

- Pati04: People Detection and
Tracking in High Resolution Panoramic Video Mosaic Patil &t al,
2004, CMU, 6p: frame differencing; face detection; color blob tracking
via mean-shift

- Peyr04: Geodesic
Remeshing Using Front Propagation Peyre' + Cohen 2004 8p: Delaunay triangulation; Voronoi diagram construction via Fast
Marching;

- Pang04:
A sentimental education: sentiment analysis using subjectivity summarization based on minimum cuts
Pang + Lee 2004 8p: text analysis via min-cut.

- Roth04:
"GrabCut": interactive foreground extraction using iterated graph cuts
Rother, Kolmogorov + Blake 2004 6p

- Roth04b: 3d
Object Modelling and Recognition Using Local Affine-Invariant Image
Descriptors and Multi-View Spatial Constraints Rothganger &al 2004
47p: difference-of-gaussians (Dog); no segmentation state; handles
high clutter levels; SIFT; SVD; RANSAC; maximal cliques;
Levenberg-Marquardt; multiple view integration; texture-reliant; slow.

- Siga04:
Tracking loose-limbed people
Sigal, Bhatia, Roth, Black, Isard 2004, 8p:
*"We pose the problem of 3D human tracking as one of inference in a graphical model"*conditional probabilities from mocap data; learned motion evolution probabilities; non-parametric belief propagation via particle filtering; continued from Siga03

- Smin04:
Generative modeling for continuous non-linearly embedded visual inference
Sminchisescu + Jepson 2004, 8p: nonlinear dimensionality reduction for
pose estimation

- Sten04: Model-Based hand Tracking Using a Hierarchical Baysesian Filter Stenger,
**thesis**, 2004, 127p

- Sten04b:
Hand pose estimation using hierarchical detection
Stenger, A Thayananthan, P Torr, R Cipolla 2004 11p: classifier
design;

- Than04: Periodic
Motion Detection and Estimation via Space-Time Sampling Thangali +
Schlaroff 2004 7p:
*"requires neither feature tracking nor object segmentation"; least squares; RANSAC;*

- Torr04:
Sharing Visual Features for Multiclass and Multiview Object
Detection Torralba, Murphy + Freeman 2004 18p: joint boosting:
cost scales "approximately logarithmically with the number of classes"

- Torr04b:
Sharing features: efficient boosting procedures for multiclass object detection
Torralba, Murphy + Freeman 2004 8p

- Urta04:
3D Human Body Tracking using Deterministic Temporal Motion Models
Urtasun + Fua 2004 15p: stereo data; PCA to avoid exponential explosion of particle
filtering;
*"good convergence properties at reduced computational cost"*

- Vice04: Robust
Object Detection in Complex Backgrounds using ICA Compression
Vicente, Fernandez, Reinoso, Perez 2004 4p: Independent Component
Analsysis (ICA) vs eigenspace analysis; FastICA; "clearly outperform
PCA" (principal component analysis);

- Viol04:
Robust Real-Time Face Detection
Viola + Jones 2004 18p: Integral Image; Adaboost; cascade. Continued
from Viol02; Riha06 speeds this up by adding CRFs (Kohl07c).

- Xiao04:
Motion Layer Extraction in the Presence of Occlusion Using Graph Cuts
Xiao + Shah 2004 8p: extracts a set of affine transformations,
multiple motion layers;.

- Xu04: Robust Real-Time Tracking of Non-rigid Objects Xu, Allen + Jin 2004 4p: color thresholding, snakes, cycle-level color conversion hacking.

- Yano04:
Finding the M Most Probable Configurations Using Loopy Belief Propagation
Yanover + Weiss 2004 8p.

- Zabi04:
Spatially coherent clustering using graph cuts
Zabih, Kolmogorov 2004 8p:
*"operates simultaneously in feature and image space"*.

- Zitn04:
High-quality video view interpolation using a layered representation
Zitnick, Kang, Uyttendaele, Winder + Szeliski 2004 9p: Bhat07 builds on this.

- Barb03:
Graph Partition by Swendsen-Wang Cuts
Barbu + Zhu 2003 8p: ergodic reversable Markov chain jumps; 20-40X
faster than DDMCMC; more general than graph cut.

- Boyk03:
Computing geodesics and minimal surfaces via graph cuts
Borykov + Kolmogorov 2003 8p:
*"graph cuts may be used to find global minimum geodesic contours"*or surfaces; Cauchy-Crofton formula; image segmentation.

- Coma03: Kernel-Based Object Tracking Comaniciu, Ramesh + Meer 2003 30p: adaptive mean-shift. Nice paper. "efficient, modular, straightforward implementation, superior performance"

- Dror03: Fragment-Based Image Completion Drori, Cohen-Or + Yshurun 2003 10p

- Tapp03: Comparison of Graph Cuts with Belief Propagations for Stereo, using Identical MRF Parameters Tappen + Freeman 2003 8p

- Eddy03:
Hidden Markov Models
Eddy 2003 5p

- Felz03: Fast Algorithms for Large-State-Space HMMs with Applications to Web Usage Analysis PF Felzenszwalb, DP Huttenlocher, JM Kleinberg 2003 8p

- Ferg03: Object
Class Recognition by Unsupervised Scale-Invariant Learning Fergus,
Perona + Zisserman 2003 8p: "flexible constellation of parts" "all
aspects of object: shape, appearance, occusion and relative scale"
"entropy-based feature detector" "expectation-maximization in a
maximimum-likelihood setting" "excellent results ... including
... cars ... and ... animals"

- Fuji03:
New Maximum Flow Algorithms by MA Orderings and Scaling
Fujishige + Isotani 2003 8p: maximum adjacency (MA). For a faster
version see Mats05

- Galu03:
Texture segmentation by multiscale aggregation of filter responses and shape elements
Galun, Sharon, Basri + Brandt 2003 8p

- Geor03:
Mean Shift Based Clustering in High Dimensions: A Texture
Classification Example
Georgescu, Shimshoni + Meer 2003 8p: Locality Sensitive Hashing (LSH),
textons

- Gold03:
Joint 3D-reconstruction and background separation in multiple views
using graph cuts
Goldlucke + Magnor 2003 6p: builds on Kolm02

- Ho03:
Clustering appearances of objects under varying illumination conditions
Ho, Yang, Lim, Lee + Kriegman 2003 8p

- Hoch03:
The Pseudoflow algorithm: A new algorithm for the maximum flow problem
Hochbaum 2003 29p

- Ishi03:
Exact Optimization for Markov Random Fields with Convex Priors
Ishikawa 2003 16p:
*generalized graph-cut to arbitrary convex cliques"*(Boyk02).

- Kolm03:
Graph based algorithms for scene reconstruction from two or more views
Kolmogorov 2003 147p (
**dissertation**

- Kolm03b:
Generalized multi-camera scene reconstruction using graph cuts
Kolmogorov, Zabih, Gortler 2003 16p

- Jin03:
Multi-view Stereo Beyond Lambert Jin, Soatto + Yezzi 2003 8p: radiance tensor

- Jone03: Fast Multi-view Face Detection
Jones + Viola 2003 11p; picks up where Viol01
left off. Handles
**rotated faces**and**profile faces**. (Levi04 needs fewer training examples.)

- Kim03: Visual Correspondence
Using Energy Minimization and Mutual Information Junhwan Kim, Vladimir Kolmogorov, Ramin Zabih 2003 8p

- Kovt03:
Partial optimal labeling search for a NP-hard subclass of (max,+) problems
Kovtun 2003 8p: Potts model; non/subconvex (max,+) problems; taking advantage of provably optimal solutions for some
pixels in NP-hard cases.

- Kuma03:
Discriminative Random Fields: A Discriminative Framework for
Contextual Interaction in Classification
Kumar + Hebert 2003 8p: DRF; neighborhood interactions in labels as
well as data; relax assumption of independence of data;
probabilistic instead of generative models;

- Kwat03:
Graphcut Textures: Image and Video Synthesis Using Graph Cuts
Kwatra &al 2003 10p:

- Levi03:
Learning How to Inpaint from Global Image Statistics
Levin, Zomet + Weiss 2003 8p: loopy belief propagation.

- Lien03:
A Detector Tree of Boosted Classifiers for Real-time Object Detection
and Tracking Lienhart, Liang + Kuranov 2003 4p

- Lin03:
Surfaces with Occlusions from Layered Stereo
Lin + Tomasi 2005 8p: continuously splined disparities; segmentation
via graph cut; gradient descent; occlusion handling;

- Liyu03: Foreground Object Detection from Videos Containing Complex Background

Liyuan Li, Weimin Huang, Irene Y.H. Gu, Qi Tian, 2003, 9p

This is the primary foreground/background discrimination algorithm used by the**OpenCV blobtracker**. Uses Rosi98 for thresholding.

- Lowe03: Distinctive Image
Features from Scale-Invariant Keypoints David G Lowe 2003 29p --
the "choo-choo" paper. Miko06 builds on this.

- McCa03:
Early results for named entity recognition with conditional random
fields, feature induction and web-enhanced lexicons
McCallum + Li 2003 4p.

- McCa03b:
Efficiently inducing features of conditional random fields
McCallum 2003 8p: max conditional log-likelihood; larger cliques; noun
phrase segmentation.

- Pari03:
A Volumetric Reconstruction Method from Multiple Calibrated Views
using Global Graph Cut Optimization
Paris, F Sillion, L Quan 2006 21p: exact global minimum; resolution to 1/10 pixel; occlusion + discontinuity handling; smoothing of graph-cut
"blockiness"; PDE result smoothing; plane separation of cameras from
scene; 30-120 CPU minutes / example.

- Pori03:
Human Body Tracking by Adaptive Background Models and Mean-Shift Analysis Porikli + Tuzel 2003
-- recommended by the OpenCV wiki for understanding the OpenCV foreground extractor.

- Rama03:
Finding and Tracking People from the Bottom Up
Ramanan + Forsyth 2003 8p:
*"builds a model of each individual"*no human dynamics model (too unreliable); loopy inference; occlusion handling.

- Ren03:
Learning a classification model for segmentation
Ren + Malik 2003 8p: superpixels; information-theoretic analysis;
gestalt cues

- Sand03:
Polynomial Time Algorithms for Network Information Flow Sanders,
Egner + Tolhuizen 2003 8p: Single source, multiple sink algorithms.

- Scha03:
The Boosting Approach to Machine Learning Schapire 2003 23p:
AdaBoost (
**book chapter**).

- Schu03: People Tracking with Anonymous and ID-Sensors Using Rao-Blackwellised Particle Filters Schulz, Fox + Hightower 2003 6p: Combining many low-grade sensor readings.

- Shak03:
Fast pose estimation with parameter-sensitive hashing
Shakhnarovich, Viola + Darrell 2003 12p.

- Sieb03: Reading People
Tracker
**thesis project**webpage Nils T Seibel, 2003

- Siga03:
Attractive people: Assembling loose-limbed models using non-parametric belief propagation
Sigal, Isard, Sigelman + Black 2003 8p: pose estimation; particle
filtering; continued in Siga04

- Smin03:
Kinematic jump processes for monocular 3D human tracking
Sminchisescu, B Triggs 2003 8p:
*"forwards/backwards flips"*;*"formal inverse kinematics solutions"*; jump-diffusion;*"complements existing methods";*particle filtering; see also Smin01,Smin03b,

- Smin03b:
Estimating Articulated Human Motion with Covariance Scaled Sampling
Sminchisescu, B Triggs 2003 24p: see also Smin01,Smin03,

- Sten03:
Filtering using a tree-based estimator
Stenger, Thayananthan, Torr + Cipolla 2003 8p: Avatars, joint angles,
hierarchical search for efficient exclusion of entire
subtrees.
*"sophisticated object detection"*(a href="Kohl07c">Kohl07c); See also Sten04.

- Varm03:
Texture Classification: Are Filter Banks Necessary?
Varma + Zisserman 2003 8p: textons. Also Varm05

- Viol03: Detecting Pedestrians Using Patterns of Motion and Appearance Viola et al, 2003, 8p

- Wain03:
Tree-reweighted
belief propagation algorithms and approximate ML estimation by
pseudo-moment matching Wainwright, Jaakkola + Willsky 2003 8p:
The original paper. Kolm05b improves the
convergence behavior.

- Wang03: Image Segmentation with Ratio Cut Wang + Siskind 2003 20p: Source code posted as a tarball.

- Wang03b: Recent developments in human motion analysis Wang, Hu + Tan 2003 34p

- Xu03:
Object Segmentation Using Graph Cuts Based Active Contours Xu,
Bansal + Ahuja 2003 8p: Improving on snakes.

- Zhao03:
Face Recognition: A Literature Survey Zhao, Challappa, Phillips +
Rosenfeld 2003 61p: available commercial systems, methods;

- Alva02: Symmetrical Dense Optical Flow Estimation with Occlusion Detection Alvarez &al 2002 15p

- Bueh02:
Minimal Surfaces for stereo
Buehler, Gortler + McMillan 2002 15p: tri-stereo in one min-cut.

- Boro02:
Pseudo-Boolean Optimization Boros + Hammer 2002 8p: Recommended by
Delong

- Boyk02:
An Experimental Comparison of Min-Cut/Max-Flow Algorithms for Energy
Minimization in Vision Boykov + Kolmogorov 2002 34p: Includes a
survey of practical applications; develops a new algorithm,
which is probably the most popular in practice as of 2008. Source code tarball available here;
Calls
**"Combinatorial Optimization"**, Cook &al 1998, Wiley,*"our favorite text-book on basic graph theory and algorithms"*.

- Brad02: Computer Vision Face
Tracking For Use in a Perceptual User Interface Gary Bradski,
Intel, 2002, 15 p: the seminal
**CamShift**paper

- Coma02: Mean Shift: A
Robust Approach toward Feature Space Analysis Comanicui Meer 2002
37p: Low-level image segmentation; One of the core meanshift papers.
*"While there are a plethora of published clustering techniques, most of them are not adequate to analyse feature spaces derived from real data"*; Opel04 claims superior results.

- Gree02:
Canny Edge Detection Tutorial Bill Green 2002. (See also the Wikipedia "Canny edge detector" article.)

- Grig02: Comparison
of Texture Features Based on Gabor Filters Grigorescu, Petkov +
Kruizinga 2002 8p: "Its convolution kernel is the product of a
Gaussian and a cosine...".

- Gust02: Particle Filters for Positioning, Navigation, and Tracking Gustafsson et al, 2002 12p

- Kolm02: Multi-camera
Scene Reconstruction via Graph Cuts Kolmogorov, Ramin + Zabih 2002 16p:
Source code tarball available here;
Gold03 builds on this.

- Lee02:
Particle filter with analytical inference for human body tracking
Lee, Cohen + Jung 2002 7p: automatic track initialization; tracking
failure recovery; reduced compute load.

- Lien02: An
extended set of Haar-like features for rapid object detection
Lienhart + Maydt, 2002, 4p: Mouth tracking;

- Numm02: A Color-Based
Particle Filter Nummiaro et al, 2002 8p: Extends Isar97: realtime tracking of nonrigid objects.

- Roy02:
Non-Uniform Hierarchical Pyramid Stereo for Large Images
Roy + Drouin 2002 8p

- Scha02:
A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence
Algorithms Scharstein + Szeliski 2002 35p: including posted source
code.

- Smin02:
Building Roadmaps of Local Minima of Visual Models
Sminchisescu + Triggs 2002, 16p: escaping local minima via transition
states; chemistry; eigenvector tracking; hypersurface sweeping;
monocular human pose estimation.

- Sudd02:
Nonparametric Belief Propagation and Facial Appearance Estimation
Sudderth, Ihler, Freeman + Willsky 2002 11p: particle filters.

- Sun02: Stereo Matching Using Belief Propagation Sun, Shum + Zheng 2002 15p

- Szel02:
Symmetric sub-pixel stereo matching
Szeliski, D Scharstein 2002 16p:

- Thay02:
Shape context and chamfer matching in cluttered scenes
Thayananthan, Stenger, Torr + Cipolla 2002 7p: Hand shapes.

- Tu02:
Image Segmentation by Data-Driven Markov Chain Monte Carlo
Tu + Zhu 2002 17p

- Viol02:
Robust Real-Time Object Detection
Viola + Jones 2002 25p: Integral Image; Adaboost; cascade. Continued
in Viol04.

- Yedi02:
Constructive Free Energy Approximations and Generalized Belief
Propagaion Algorithms Yedidia, Freeman + Weiss, 2002 23p:
continuation of Yedi00; factor graphs instead of
Markov Random Fields; "region graph method is
the most general"; "GBP algorithms often dramatically outperform BP algorithms"

- Zale02:
Fast algorithms of Bayesian Segmentation of Images
Zalesky 2002 13p:
*"very fast special case"*(Kolm03)

- Zhu02:
What are Textons?
Zhu, Guo, Wang + Xu 2002 15p

- Arul01: A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking

Arulampalam &t al, 2001, 15p: This describes the**OpenCV blobtracker**algorithm used to track blobs.

- Berg01:
Geometric blur for template matching
Berg + Malik 2001 8p.

- Boyk01:
Interactive Graph Cuts for Optimal Boundary & Region Segmentation of
Objects in NDImages Boykov + Jolly, 2001, 8p

- Boyk01b:
A New Algorithm for Energy Minimization with Discontinuities
Boykov, Veksler + Zabih, 2001, 16p

- Coma01: The Variable Bandwidth
Mean Shift and Data-Driven Scale Selection Comaniciu, Ramesh +
Meer 2001: adaptive mean shift.

- Deut01:
Automatic Partitioning of High Dimensional Search Spaces associated
with Articulated Body Motion Capture
Deutscher, Davison + Reid 2001.

- Fuen01: People tracking
in surveillance applications Fuentes + Velastin 2001 6p: multiple
objects; blob merging and splitting; no background model;

- Hoch01:
An efficient algorithm for image segmentation, Markov Random Fields
and related problems.
Hochbaum 2001 16p

- Iwat01:
A Combinatorial Strongly Polynomial Algorithm for Minimizing
Submodular Functions
Iwata, Fleischer, Fujishige 2001 21p.

- Kade01: An Improved Adaptive Background Mixture Model for Real-Time Tracking with Shadow Detection P. KadewTraKuPong and R. Bowden 2001 5p: The mixture-of-gaussians (mog) paper which is the basis for the
**OpenCV blobtracker**mog module. Also addresses shadow removal.

- Kang01:
Handling occlusions in dense multi-view stereo
Kang, Szeliski + Chai 2001 35p: Vulnerable to local minima; assymetric
image handling; Kolm03 builds on this.

- Kolm01: Computing Visual
Correspondence with Occlusions via Graph Cuts Vladimir Kolmogorov,
Ramin Zabih 2001: graph cut, "expansion move"; stereo matching;
Source code tarball available here;

- Laff01:
Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data
Lafferty, McCallum + Pereira 2001 8p.

- Leun01:
Representing and Recognizing the Visual Appearance of Materials using Three-dimensional Textons
Leung, Malik 2001 16p

- Matu01:
Polyhedral Visual Hulls for Real-Time Rendering
Matusik, Buehler + McMillan 2001 12p.

- Seni01: Appearance Models for Occlusion Handling

Andrew Senior &t al, 2001, 8p.

This describes an alternate blob-tracking algorithm, not currently implemented in OpenCV.

- Shar01:
Segmentation and boundary detection using multiscale intensity measurements
Sharon, Brandt + Basri 2001, 8p:
combining contrast, texture and edge info; SWA; coarse plus fine
properties; weak edges;

- Smin01:
Covariance scaled sampling for monocular 3D body tracking
Sminchisescu, B Triggs 2001 8p: >=30 joint parameters;
*"combines robust optical flow, edge energy and motion boundaries"*; limitations of CONDENSATION; see also Smin03

- Viol01:
Rapid Object Detection using a Boosted Cascade of Simple Features Viola
+ Jones 2001 9p: Integral images, Haar-like features; AdaBoost. This
work is continued in Jone03; "highly
successful"(Wolf05) "first highly accurate as
well as real-time frontal face detector" (Levi04)

- Acha00: US Patent 7027649 - Computing the Euler number of a binary image

: By computing runs, then run adjacencies. - Amen00: Accurate and Efficient Unions of Balls

: Balls; Medial Axis Transform; Voronoi diagram; fast collision detection; . - Boug00: Pyramidal Implementation of the Lucas Kanade Feature Tracker: Description of the algorithm Jean-Yves Bouget, 2000, Intel, 9p

- Coma00: Real-Time Tracking of
Non-Rigid Objects using Mean Shift Comaniciu et al 2000 8p
(
**best paper award**, patent): Bhattacharyya coefficient; football;

- Felz00:
Efficient Matching of Pictorial Structures
Felzenszwalb + Huttenlocher 2000 10p: See also Kuma05, Kuma04, Felz05

- Free00:
Learning Low-Level Vision Freeman, Pasztor + Carmichael 2000 8p:
The first application of belief propagation to computer vision (Szel06); detailed description of
Loopy Belief Propagation

- Iwat00:
A Combinatorial Strongly Polynomial Algorithm for Minimizing Submodular Functions
Iwata, Fleischer + Fujishige 2000 21p

- Kim00:
Incorporating spatial priors into an information theoretic approach
for fMRI data analysis
Kim, Fisher, Tai, Wible + Willsky 2000 10p: Image segmentation via graph-cut

- Kutu00:
A Theory of Shape by Space Carving
Kutulakos + Seitz 2000 8p: n-view reconstruction; photo hull; plane
sweeps; scene radiance; occlusion handling; Vedu00 and Boyk06b build on this.

- Leve00:
Statistical shape influence in geodesic active contours
Leventon, Grimson + Faugeras 2000 8p: Principal Component Analysis (PCA)

- Lewi00: Direct Search Methods: Then And Now Robert Michael Lewis, Virginia Torczon, Michael W. Trosset 2000 16p: nonlinear optimization; quasi-Newton methods; Nelder-Mead simplex search;

- Lou00: Semantic Interpretation of Object Activities in a Surveillance System Lou, Liu, Tan & Hu, 2000, 4p

- Maso00:
Boosting Algorithms as Gradient Descent Mason, Baxter, Bartlett +
Frean 2000 7p: Recommended by Delong

- Merw00: Unscented Particle Filter Merwe, Doucet, Freitas + Wan 2000 7p.

- Para00:
Geodesic Active Contours and Level Sets for the Detection and Tracking of Moving Objects
Paragios + Deriche 2000 15p

- Roy00:
MRF solutions for probabilistic optical flow formulations Roy + Govindu 2000 7p.

- Schn00: A Statistical Method for 3D Object Detection Applied to Faces and Cars Schneiderman & Kanade, 2000, 6p

- Scho00:
Video textures
Schodl, Szeliski, Salesin + Essa 2000, 6p

- Schr00:
A combinatorial algorithm minimizing submodular functions in strongly polynomial time
Schrijver 2000, 8p

- Shi00:
Normalized Cuts and Image Segmentation Shi + Malik 2000 6p: Low-level image segmentation.

- Side00:
A framework for modeling the appearance of 3D articulated figures
Sidenbladh, Torre + Black 2000 8p: linear subspaces; particle filtering.

- Side00b:
Stochastic tracking of 3D human figures using 2D image motion
Sidenbladh, Black + Fleet 2000 17p: generative model; graylevel difference likelihood;
pose and joint angle prior; particle filtering; optical flow;
perspective model; limb self occlusion; complex background removal;
frame-to-frame brightness constancy.

- Snow00:
Exact Voxel Occumancy with Graph Cuts Snow, Viola + Zabih 2000 8p.

- Stau00: Learning Patterns of Activity Using Real-Time Tracking Stauffer + Grimson 2000 11p: Reference example of implementing mixture-of-gaussians (mog) background subtraction.

- Sump00: Learning
Spatio-Temporal Patterns for Predicting Object Behavior Sumpter &
Bulpitt, 2000 10p: neural nets;

- Vedu00:
Shape and motion carving in 6 D
Vedula, Baker,Seitz + Kanade 2000 7p: Builds on Kutu00.

- Veks00:
Image segmentation by nested cuts
Veksler 2000 6p

- Yedi00: Generalized
Belief Propagation Yedidia, Freeman + Weiss, 2000 9p:
"significantly more accurate than ordinary BP" picked up in Yedi02

- Birc99:
Multiway cut for stereo and motion with slanted surfaces
Birchfield + Tomasi 1999 7p: alternativing region finding with region
affine parameter finding.

- Bobi99:
Large Occlusion Stereo
Bobick + Intille 1999 20p: First paper to handle stereo occlusions (Boyk02)

- Boyk99:
Fast Approximate Energy Minimization via Graph Cuts Yuri Boykov,
Olga Veksler, Ramin Zabih 1999 41p: US Patent
6744923; one of the two most popular graph
cut solution algorithms (Szel06); studies
size-two cliques, Kohl07cKohl07c extends
this to larger cliques and more general energy functions.

- Boyk99b:
A New Bayesian Framework for Object Recognition
Boykov + Huttenlocher 1999 7p

- Cohe99:
A Simple, Fast and Effective Rule Learner Cohen + Singer 1999 8p:
SLIPPER; using AdaBoost to generate human-comprehensible rule sets

- Coma99: Mean Shift Analysis and Applications
- Cutl99: Robust Real-Time Periodic Motion Detection, Analysis, and Applications Cutler + Davis 1999 35p

- Elga99: Non-parametric Model for Background Subraction Elgammal, Harwood + Davis 1999 17p: Kernel methods, shadow suppression. Generalization of mixture-of-gaussians (mog).

- Freu99:
A Short Introduction to Boosting Freund + Schapire 1999.

- Gavr99: The Visual Analysis of Human Movement: A Survey Gavrila 1999 43p

- Gion99:
Similarity Search in High Dimensions via Hashing
Gionis, Indyk + Motwani: the seminal paper on Locality Sensitive Hashing.

- Karg99:
Random Sampling in Cut, Flow, and Network Design Problems
Karger 1999 10.

- Klei99:
Approximation Algorithms for Classification Problems with
Pairwise Relationships: Metric Labeling and Markov Random Fields
Kleinberg, Tardos 1999 24p:
*"generalization of multiway cut"*;*"first nontrivial polynomial-time approximation algorithms for..."*.

- Leun99:
Recognizing Surfaces Using Three-Dimensional Textons
Leung, J Malik 1999 8p

- Lowe99: Object
Recognition from Local Scale-Invariant Features Lowe 1999 8p: The
seminal SIFT (Scale-Invariant Feature Transform) paper: "similar to neuron in the inferior temporal cortex
used for object recognition".

- McKe99:
Tracking colour objects using adaptive mixture models
McKenna, Raja + Gong 1999 7p

- Meer99: Robust Techniques for Computer Vision Peter Meer 1999 78p (
**book chapter**): "Why nonparametric methods?"; elements of a model; functional vs structural models; robust ellipse fitting; large gross error sensitivity in LMedS and RANSAC; regression vs location problems; expectation maximization; adaptive mean shift; pbM-estimator (=="projection-based M-estimator"); LTS estimator; "MLESAC and MSAC ... are superior to RANSAC";

- Mihc99:
Low-complexity image denoising based on statistical modeling of
wavelet coefficients
MÄ±hcak, I Kozintsev, K Ramchandran, P Moulin 1999, 9p

- Nevi99: Face Recognition Using an Embedded HMM Nevian and Hayes, 1999, 6 p

- Ouha99: 3D Hand Gesture Tracking by Model Registration Ouhaddi + Horain 1999 4p

- Page99:
Nonparametric Markov Random Field Models for Natural Texture Images
Paget 1999 234p:
**dissertation**

- Roy99:
Stereo Without Epipolar Lines: A Maximum-Flow Formulation Roy 1999
15p:

- Seit99:
Photorealistic Scene Reconstruction by Voxel Coloring
Seitz + Dyer 1999 32p.

- Stau99:
Adaptive background mixture models for real-time tracking
Stauffer, WEL Grimson 1999 6p: See also Kade01;
Kohl07c builds on this.

- Stew99: Robust Parameter Estimation in Computer Vision Stewart 1999 32p (survey): M-estimators; least-median of squares; Hough transforms; RANSAC; "fundamental matrix" (one scene, two viewpoints); SIAM Review

- Stor99:
Skin colour detection under changing lighting conditions
Storring, Andersen + Granum 1999 8p

- Veks99: Efficient
graph-based minimization methods in computer vision Veksler 1999
**thesis**: jump-move solution of graph cut problems.**This thesis provides an excellent introduction to graph cut techniques in computer vision.**

- Szel99:
An experimental comparison of stereo algorithms Szeliski + Zabih 1999
17p

- Birc98:
A Pixel Dissimilarity Measure That Is Insensitive to Image Sampling
Birchfield + Tomasi 1998 6p: a basic technique widely used.

- Birc98b:
Depth Discontinuities by Pixel-to-Pixel Stereo
Birchfield + Tomasi 1998 8p.

- Boyk98:
Markov Random Fields with Efficient Approximations
Boykov, Veksler + Zabih 1998 8p: "swap move" algorithm, one of the two most
popular graph-cut solution algorithms (Szel06).
"Expansion move" (Boyk99) is apparently strictly better.
Source code tarball available here;

- Boyk98b:
A Variable Window Approach to Early Vision
Boykov, Veksler + Zabih 1998 22p

- Frie98:
Additive Logistic Regression: a Statistical View of Boosting
Friedman, Hastie + Tibshirani 1998 43p: the seminal
**gentleBoost**paper.

- Hofm98:
Unsupervised Texture Segmentation in a Deterministic Annealing Framework
Hofmann, Puzicha + Buhmann 1998 48p.

- Ishi98:
Occlusions, Discontinuities, and Epipolar Lines in Stereo
Ishikawa + Geiger 1998 14p:
*First graph-cut method to elegantly handle occlusions -- but Kolm01 does it better*(Boyk02); one of the first papers to combine MRF with shape prior (Kohl07c).

- Ishi98b:
Segmentation by Grouping Junctions
Ishikawa + Geiger 1998 6p: one of the first maximum-flow image
segmentation papers.

- Page98:
Texture synthesis via a noncausal nonparametric multiscale Markov random field
Paget 1998 9p: Higher order markov fields.

- Pigh98: Synthesizing realistic facial expressions from photographs Pighin et al, UW, 1998, 10p

- Pyly98: Is Vision Continuous with Cognition? The Case for Cognitive Impenetrability of Visual Perception Pylyshyn 1998 50p

- Rosi98: Thresholding for Change Detection Rosin 1998 6p: This is the thresholding technque used by the
**OpenCV blobtracker**mm2003 histogram-based background model: see Liyu03.*"Most promising were the spatial methods. Both the Poisson noise model and the stable Euler number reliably gave good results..."*

- Roy98:
A maximum-flow formulation of the N-camera stereo
correspondenceproblem Roy + Cox 1998 8p:
*first use of graph cut for stereo reconstruction*(Boyk02).

- Scha98:
Improved Boosting Algorithms Using Confidence-rated Predictions Schapire + Singer 1998 4p: Seminal Real AdaBoost paper.

- Smyt98:
Belief Networks, Hidden Markov Models, and Markov Random Fields: a
Unifying View Smyth 1998 11p UC Irvine.

- Torr98: Robust Computation and Parametrization of Multiple View Relations Torr + Zisserman 1998 6p: Seminal paper introducing MSAC, MLESAC (Maximum Likelihood). Claims MSAC strictly dominates RANSAC: Better at no extra compute cost. Fundamental matrix, planar homography; epipolar constraint

- Chek97:
Experimental Study of Minimum Cut Alorithms Chekuri &al 1997(?)
10p: see also the full
version (138p) and source code.

- Cher97:
On Implementing the Push-Relabel Method for the Maximum Flow Problem
Cherkassky + Goldberg 1997 18p.

- Frie97: Image segmentation in video sequences: A probabilistic approach Friedman + Russell 1997 13p UCB: Cited as -the- seminal mixture-of-gaussians (mog) background subraction paper. "Sufficient statistics." Roadwatch traffic project.

- Gold97:
Length functions for flow computations Goldberg + Rao 1997 19p:
*"unlikely to be efficient in practice because it has substantial overhead ... it is very likely that the empirical running time will be near the theoretical worst-case"*(Rico05)

- Isar97: CONDENSATION
-- Conditional Density Propagation for Visual Tracking Isard +
Blake 1997 24p: One of the first particle-filtering
papers. ("CONDENSATION" is the algorithm name.) Very nice work --
worth re-reading. Numm02 builds on this; Smin01 critiques it.

- Juli97: A new Extension of the Kalman Filter to Nonlinear Systems Julier + Uhlmann 1997 12p: The seminal Unscented Kalman Filter paper. "[The Extended Kalman Filter] is difficult to implement, difficult to tune, and only reliable for ...".

- Szel97:
Creating full view panoramic image mosaics and environment maps
Szeliski, Shum 1997 8p:
*"hand-held digital camera images can be stitched seamlessly"*;*"fast and robust"*;*"directly recovers 3D rotations"*.

- Torr97: MLESAC: A new robust estimator with application to estimating image geometry Torr + Zisserman 1997 19p: "MLESAC" == "Maximum Likelihood Estimation SAmple Consensus".

- Wolf97: Geometric
Hashing: An Overview Wolfson + Rigoutsos 1997 12p.

- Wren97: Pfinder: Real-Time Tracking of the Human Body Wren &al 1997 6p: I believe this is one of the seminal mixture-of-gaussians (mog) background model papers.

- Boum96:
A Multiscale Random Field Model for Bayesian Image Segmentation Bouman 1996 43p

- Brei96:
Bagging Predictors Breiman 1996 20p

- Gavr96:
3-D model-based tracking of humans in action: a multi-view approach
Gavrila, Davis 1996 8p: chamfer matching

- Karg96:
A New Approach to the Minimum Cut Problem
Karger + Stein 1996 37p: randomized algorithm

- Cheu95: Mean Shift, Mode Seeking and Clustering Cheng, 1995, 10p: "re-ignited interest" in mean-shift analysis (Comaniciu + Meer 2002); k-means a limit case.

- Freu95:
A decision-theoretic generalization of on-line learning and an
application to boosting Freund + Schapire, 1995, 35p: the seminal
**AdaBoost**paper.

- John95: Learning the Distribution of Object Trajectories for Event Recognition Johnson & Hogg 1995 10p

- Rosi95: Image difference threshold strategies and shadow detection Rosin + Ellis, 1995, 10p: Using Euler number to measure connectivity as a cheaper alternative to counting regions (cited by Rosi98; image registration via cross-correlation minimization;

- Hao94:
A Faster Algorithm for Finding the Minimum Cut in a Directed Graph Hao, 1994 23p.

- Naga94:
Implementing an efficient minimum capacity cut algorithm
Nagamochi, Ono + Ibaraki, 1994 20p: one of the papers Chek97 builds upon..

- Stoe94: A
Simple Min-Cut Algorithm Stoer, 1994 7p: One of the first modern
efficient min-cut algorithms, recommended by The Algorithm Design Manual,
which also cites
*Network Flows*by Ahuja, Magnanti + Orlin 1993 Prentice Hall as the definitive reference.

- Wu93:
An Optimal Graph Theoretic Approach to Data Clustering: Theory and Its
Application to Image Segmentation Wu + Leahy 1993 13p: One of the
applications of graph-cut algorithms to computer vision.

- Chia92:
Dynamic algorithms in computational geometry
Chiang + Tamassia 1992 48p.

- Rigo92:
Massively Parallel Bayesian Object Recognition
Rigoutsos 1992 221p:
**dissertation**: Geometric hashing on the connection machine.

- Cohe91:
Dynamic expression trees and their applications
Cohen + Tamassia 1991 25p.

- Swai91:
Color indexing
Swain + Ballard 1991 22p:
*"color histograms are stable object representations in the presence of occlusion and over change in view"*(McKe99 argues for gaussian mixture models instead.)

- Szel91:
Surface Modeling with Oriented Particle Systems
Szeliski, Tonnesen 1991 38p.

- Rabi89: A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition Rabiner, 1989 30p: ergodic vs left-right HMMs; continuous observation densities; null transitions; maximum likelihood (ML) maximum mutual information (MMI) and minimum discrimination information (MDI) optimization; implementation issues; coping with insufficient training data; limitations.

- Gold88:
A New Approach to the Maximum-Flow Problem Goldberg +
**Tarjan**, 1988 20p:*"The first push-relabel algorithms ... offer the most robust performance for general graphs ... surpassed by [Boyk02] for ... 2D grids"*(Delo08). Delo08 bases a highly efficient and scalable grid-graph solution method on this.

- Fred87:
Fibonacci heaps and their uses in improved network optimization algorithms
Fredman +
**Tarjan**, 1987 20p: needed by (e.g.) Stoe94 .

- Pear86:
Fusion, Propagation, and Structuring in Belief Networks
Pearl 1987 48p.

- Suzu85: Topological
structural analysis of digital binary image by border following
Suzuki + Abe, 1985: Source of the contour extraction algorithm used by
**OpenCV**; paper does not appear to be available online.

- Gema84:
Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images
Geman + Geman 1985: One of the classics of the field.

- Torr84:
On Edge Detection
Torre + Poggio 1984 41p.

- Slea83:
A data structure for dynamic trees
Sleator +
**Tarjan**1983 30p.

- Horn81:
Determining Optical Flow
Horn + Schunck 1981 19p.

- Fisc80: Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography Fischler + Bollers 1980 15p: The seminal RANSAC paper, the standard computer vision regression algorithm. Why least-squares isn't enough. Landmark Determination Problem (LDP -- deducing camera position).

Back to Cynbe's Linux Computer Vision App.

Cynbe ru Taren Last modified: Mon Jan 26 03:27:39 CST 2009