Change detection in epilepsy monitoring video based on Markov random field theory

Video recording is often conducted during electroencephalographic (EEG) recording from epilepsy patients. This type of video contains small local motions and mostly idle segments. Change detection provides a powerful tool to detect local motions for analyzing, editing and archiving this type of video. Classic change detection methods utilize predetermined thresholds to test variations between frames. The determination of these thresholds, however, is often problematic. In this work, we present a new approach to change detection from the probabilistic optimization point of view. By modeling images as Markov random fields, we formulate change detection into a problem of seeking the optimal configuration of the change detection map (CDM). An algorithm that searches the optimal configuration is constructed by applying mean field theory (MFT), which greatly reduces computational complexity. Our experimental results show that this method detects changes accurately and is resilient to noise.