Detecting moving shadows: algorithms and evaluation

Making Background Subtraction Robust to Sudden Illumination Changes

“…This problem can be alleviated by explicitly detecting the shadows [8]. Most of them consider them as binary [8], with the notable exception of [9] that also considers penumbra by using the ratio between two images of a planar background. Our approach also relies on image ratios, but treats shadows as a particular illumination effect, a wider class that also include the possibility of switching lights on.…”

A comparison of two Monte Carlo algorithms for 3D vehicle trajectory reconstruction in roundabouts

“…1), typically extracted by background subtraction [4] or frame-by-frame difference [21]. To estimate the vehicle pose, some region-based algorithms minimize a metric, as for the edgebased case [22], while other algorithms calculate a convenient score for a set of hypothesized model poses [23,8].…”

“…Most of the existing visual vehicle tracking systems propose a 2D approach (2D tracking hereafter): these systems identify moving vehicles on the image plane, e.g., by identifying their blobs (see Fig. 1) via background subtraction [4], and they track their trajectories on this plane [5,2]. Although, in some applications, this type of estimate might be sufficient to fully understand the vehicle behavior, in many cases, especially in roundabout intersections, we need to estimate vehicle trajectories with high (a) Vehicle.…”

Shining a Light on Human Pose: On Shadows, Shading and the Estimation of Pose and Shape

“…Interestingly, most work in shadow detection has focused on removing the shadows to improve foreground segmentation [9]. Distinguishing foreground from shadows can be challenging since both differ significantly from the background.…”