2008-IUPR-17Mar_1723.pdf

0.2 100 100-EER 1000 104 1124 1137 160 1614 1627 17.61 18.22 1997 1999 2.4 200 2000 2002 2004 2005 2006 2007 2330 2340 255 261 269 274 2nd 300 3099 3104 4.1 4.2 40.51 400 44.5 48.0 497 5.02 5.31 5.41 52.04 53.6 54.87 6.25 6.99 60.01 600 628 63.23 63.56 641 66.3 67.20 67.4 68.3 69.76 71.41 72.66 747 751 757 767 780 785 791 8-connected 8.09 800 83.87 831 843 85.1 9.54 91.1 961-386261 979 988 A.M ACCV Abstract Activity Adrian Algorithms Also Analysis Appl Applications Approach Approximation Artificial Autonomous Azarbayejani BMVC Background Backgrounds Based Basis Bayesian Bernier Bilayer Bischof Body Boolean Both Boykov Breuel1 Brumitt C.R CRF CVPR Camoufl Camouflage Campus Center Change Changing Cheung Classification Classifiers Closest Co-Occurrence Color Comaniciu Compared Comparison Complex Component Computer Conditional Conventional Criminisi Culibkr Culibrk Curtain Cut DFKI Darrell Dataset Datasets Davis Density Department Detection Discriminative Discussion Duda Due ECCV EER EURASIP Each Edition Eigenbackgrounds Eigenspace Elgammal Energy Equation Error Essa Eval Experiment Experimental Experimenting Experiments Field Fig Figure Finally For Foreground Fountain Framerate From Function Furht Further GHz Gaussian Gaussians German Gibbs Given Grabner Graph Graphcut Grimson Ground Group H.-Y Han Hart Harwood Here Hidden Histograms Horprasert Huang Human I.A ICCV ICSMC IEEE III Illumination Image Images InViRe Influence Innovation Instead Int Integration Intelligence Interactions Interscience Introduction Invariant J.M Kaiserslautern Kalva Kamath Kernel Kolmogorov Krumm L.S Learning Left Local MAP MLP MOG Maintenance Marques Max-Flow Methods Meysers Min-Cut Minimization Mode Model Modeling Modelling More Most Motion-based Multilayer Netw Network Neur Neural Non-parametric Noriega Note Number ONLINE Object Office Oliver Once Opteron Other Our Overall P.M PAMI PCA Pattern Patterns Pentland Perceptron Perf Pfinder Piccardi Practice Principal Principles Proc Propagation Publications Quality Quantitative RBF Radial Random Rate Real Real-Time Realtime Recently Recognition References Related Research Result Review Rheinland-Pfalz Right Robust Rosario Roth Row SMC Sample Samples Science Second Section See Segmentation Sequence Sequences Sequential Setup Shadow Shum Sign Since Socek Some Starting Statistical Statistics Stauffer Stiftung Stork Subtraction Such Sun Surv Synthesized System Table Tang Technical Techniques The These This Thomas Those Tian Time Toyama Track Tracking Traffic Training Trans Tree-based Two Ulges1 Understanding Units University Urban Using Validation Via Video Virtual Vision Vol W.Surf W.Trees WACV Wallflower WaterSurf.Curtain WaterSurface Wav.Trees WavingTrees WavingTress When While Wiley Winn Work Workshop Wren Yin Zhang abrupt absence according achieve achieved acquired actually adapt adapted adaption added additional additive adequate algorithm algorithms allows alternative analysis analyze approach approaches approximately architecture arg art assessment assign associated assumed assumes attenuation available background background-only backpropagation based baseline beginning belongs benchmark benchmarks best better bias boosting boundaries boundary brightness camera capable careful carried case cast caused changes characteristics choose chosen chroma class classes classification classifier classifiers clique close co-occurrence color colors combined comparable comparably compare compared compares comparisons competitive complex components compute conditions consisting consists constant constraint constraints construct constructed contrast contributions control conventional converges copied core correction correctly corresponding corresponds corrupted covered covering cues cut dark data dataf dataset dataset3 datasets days decide decided deciding decision decomposition define defined degree demanded demonstrate demonstrated demonstrating denoted denoting densities density depends described description detecting determined determines deviation difference different difficult difficulties direction directions directly discriminative distance distinction distribution distributions does downloads downscaling drops e-U e.g easier effect efficiently effort energy environment epochs equal error error-prone errors estimate estimated estimates evaluations evidence exist expected experience experiment experiments extracted fact fail fails fair far fashion feature features fed field finally fit fitting fixed flickering focal follow followed following follows foreground form formulation fps frame frames framework frequently fully function fused gain generative given gives global good grad gradient graph gray ground hand hand-tuned hard held heuristic hidden highlighted histograms http hypothesis i.e idea illumination illustrate illustrated illustration image images implementation implemented improved improvement improves in-depth increases increasing influence informatik.uni-kl.de information input integrate integrated integrates integrating integration intensity interesting internal interpretation iupr.dfki.de kept kernel key known label lack large later layer leads learn learning length level light lighting like linearly literature local locally log logarithm longer low-dimensional maintain majority makes manner manually mask masks max maximizes measure measures method methods min minimizes minimizing misclassifications mixture-of-Gaussians mixtures model modeled modelling models monitor motion moving multiple near-realtime need needed negligible neighbor network neural noise noisy non-background non-parametric note number object objects observation observations observed obtain obtained obviously occur occurring occurs office offline ones online open optimal optimization optimizes optimum outlined outperformed overall overcome overfitting overruled paired pairs paper parallelized parameter parameters parametric partially particular parts patch patterns penalized perform performance performs phase phenomena pictures pixel pixel-wise pixels plain plant plots plotted poor popular position posterior posteriors potentials practical presence present presented previous previously prior probabilistic probability problem problems processor project projected projection projects pronounced proper properly properties propose proposed prototype proved proven provided public publications publicly purpose qualitative quality quantitative quantitatively question radius random randomly rate rates reached react reacts real realistic really receptive red referred region regions related relative reliably replacement reported represent representing require requires res resolution rest result resulting results returned revealed reviews robust robustly robustness row rules run running runs sample sampled samples sampling satisfy saturation scaled scales scene scene4 scenes scores sec seen segmentation segmentations segmented segmenters segmenting self-created self-generated sensitive separate separately separating sequence sequences set setup shadow shadows shake short showing shown shows significant significantly similar simple simply simulates situations slight slightly small smooth smoothness so-called solution sources space squared standard start state-of-the state-of-the-art static statistically stick stored strategy strength stresses strong study subsampled subtraction sudden suffer suggested suitable supervised supported surfaces surrounding switched switches synthesize synthesized synthesizing synthetic systems t-test taken takes taking targeted technique techniques tend term test tested testing texture thoroughly threshold thresholding thresholds time tmb train trained training treating treebased trees truly truth tune tuned tuning types ulges uniform units unoptimized unsupervised update updated updates use used using value values variables variance varying vectors view virtual water waving way ways weather weighs weighted weights width work worse www.iupr.org