资 源 简 介
Robustnesstochangesinilluminationconditionsaswellas viewing perspectives is an important requirement formany computer vision applications. One of the key fac-ors in enhancing the robustness of dynamic scene analy-sis that of accurate and reliable means for shadow de-ection. Shadowdetectioniscriticalforcorrectobjectde-ection in image sequences. Many algorithms have beenproposed in the literature that deal with shadows. How-ever,acomparativeevaluationoftheexistingapproachesisstill lacking. In this paper, the full range of problems un-derlyingtheshadowdetectionareidenti?edanddiscussed.Weclassifytheproposedsolutionstothisproblemusingaaxonomyoffourmainclasses, calleddeterministicmodeland non-model based and statistical parametric and non-parametric. Novelquantitative(detectionanddiscrimina-ionaccuracy)andqualitativemetrics(sceneandobjectin-dependence,?exibilitytoshadowsituationsandrobustnesso noise) are proposed to evaluate these classes of algo-rithms on a benchmark suite of indoor and outdoor videosequences.
