Occlusion detection in video sequences

Abstract

An occlusion is the region between two overlapping objects with disparate motion. Detecting these occluded objects is crucial for many of the video processing. The Occlusion detection is decomposed into two independent sub problems. The First is to detect foreground objects on a frame-wise basis, by labeling each pixel in an image frame as either foreground or background. The second is to couple object observations at different points in a sequence to yield the object's motion trajectory and Occlusion. The motion segmentation is based on an adaptive background subtraction method that models each pixel as mixture of Gaussians. The Gaussian distributions are then evaluated to determine which are most likely to result from a background process. This is useful to track moving objects and detect occlusion in lighting changes, repetitive motions from cluster, and long term scene changes.