Network Video Frame-Skip Modeling and Simulation

A high video frame rate is ideal for many computer vision applications; however, this is not is usually the case for video feeds streamed from network cameras, which also exhibit frame skipping. Frame skipping can be considered as a type of degradation of the live-streamed video, and is detrimental to object tracking and other computer vision tasks. In this paper, we empirically model frame skipping and present a method for simulating frame skipping using videos with high frame rates. This simulation method was then validated by comparing the frame rate histograms of real and simulated frame-skipped videos. Finally, we demonstrate how frame skipping degrades the performance of an object tracker.