Evaluation of background subtraction algorithms for object extraction

There is an increase in need of video surveillance applications. Intelligent video surveillance (IVS) includes public safety and security applications, including authenticity control, crowd flow direction and crowd analysis, human behaviour detection and analysis etc. The critical part of IVS system is proper foreground estimation using background subtraction algorithms. This is a challenging task due to variations in illumination, background motion due cluttering noise like swaying trees, flowing water, etc. and the slow moving objects introduce noise in the background estimated. We have precisely concentrated on such challenges. The purpose of this evaluation is to give an overview and categorization of the approaches based on the performance measures like Precision, Recall, F measures (F1), Similarity, Matching Index and Average Classification Error. And also the available techniques are compared based on the computational complexity parameters in terms of Big-O along with their limitations for the improvement in the efficiency of the background subtraction algorithms.