A new change detection method and its application to epilepsy monitoring video

Video recording is often conducted during epilepsy monitoring. This video provides important information correlating electrical activity with a patient´s behavior during a seizure. Change detection in this type of video provides a tool to detect the motion of the patient to aid in diagnosis. Traditional methods detect changes by testing variations between frame pairs. Although effective in common applications, these methods are not satisfactory in detecting movements in medical video involving small local motion. In this work, we present a new change detection method by utilizing multiple video frames. Temporal domain properties are explored to warrant more robustness in the detection of small changes. The effectiveness of the new method is demonstrated by experimental results.