ABSTRACT
The detection of moving objects is the fundamental pre-processing step in many computer vision and image processing applications, such as video inpainting, compression, privacy, surveillance, segmentation, optical flow, and augmented reality [1], [14]. This basic step requires an accurate and efficient background subtraction (also known as foreground detection). Typically, the background subtraction process consists of isolating the moving objects called “foreground” from the static scene called “background.” However, it becomes a really hard task when the background scene contains more variations such as sudden global illumination changes, waving trees, water surface, moving curtains, bootstrapping, etc. In addition, color saturation and bad weather conditions are major well-known background modeling challenges.
