Efficient Incorporation of Optical Flow into Visual Motion Estimation in Tracking

Abstract

Recent developments in digital technology have increased acquisition of digital video data, which in turn have led to more applications in video processing. Video sequences provide additional information about how scenes and objects change over time when compared to still images. The problem of tracking moving objects remains of great research interest in computer vision on account of various applications in video surveillance, monitoring, robotics, and video coding. For instance, MPEG-4 video standard introduced video object plane concept, and a decomposition of sequences into object planes with different motion parameters [1]. Video surveillance systems are needed in traffic and highway monitoring, in law enforcement and security applications by banks, stores, and parking lots. Algorithms for extracting and tracking over time moving objects in a video sequence are hence of importance.