Illumination-invariant change detection model for patient monitoring video

Video recording is often conducted in the medical environment. Change detection provides a powerful tool to detect dynamic changes in the video to aid in monitoring and diagnosis. Illumination variation presents a typical problem for a change detection method to gain robustness. In this work, we describe a new method based on an illumination model and test statistics to reduce the sensitivity of detection to illumination changes. The effectiveness of this method is demonstrated by our experimental results.