Subsense is a universal change detection method for foreground/background segmentation in video sequences. It relies on spatiotemporal binary features and color information to detect changes, allowing camouflaged foreground objects to be detected more easily while ignoring most illumination variations. Subsense uses pixel-level feedback loops to dynamically adjust its internal parameters without user intervention, based on monitoring model fidelity and local segmentation noise levels. This enables it to outperform 32 state-of-the-art methods on standard datasets, achieving real-time processing speed on a desktop CPU using its available C++ implementation.