Modified LBP-Based Background Modeling for Video Object Detection.

The paper proposes a background modeling-based approach for the detection of foreground objects. Often, background has dominant textural behavior with some of the dynamic entities. In order to take care of both, the paper proposes two variants of Local Binary Pattern (LBP), LBP-A and LBP-EA, for background modeling. The histograms of the proposed variants of LBP learn the new video sequence to model the complex background and accurately differentiate the foreground from the background. The proposed model learning approaches have successfully been tested with different frames of PETS sequences and the efficacy of the proposed models has been found to be better than that of LBP-based modeling. [ABSTRACT FROM AUTHOR]