On the Development of Foreground Detection under Complex Background

Foreground detection is a prime task in the field of computer vision for targeting the emerging applications like video surveillance, object tracking, action recognition, scene analysis. For moving object detection, it is always desirable to accurately extract the foreground under complex background conditions with less computational overhead. In this work, we propose a multifeature-based moving object detection scheme, where the feature vector for each pixel constitutes gray level intensity value and extended scale-invariant local ternary pattern (E-SILTP) over a local region. Further, to improve the detection accuracy with minimum computational cost, extended Canberra distance is employed for similarity distance between model and current pixel instead of popular Mahalanobis distance and Forstner distance. The experimental results are validated using some standard data sets and shows superior performance than that of the benchmark schemes.