A detection method and a detection device for detection of at least one region of interest (ROI) using the same are provided. A plurality of successive frames is captured by an image sensor. A first frame among the plurality of successive frames is divided into a plurality of sub regions. A first vital-sign feature of a first sub region among the plurality of sub regions is obtained. A first feature signal is generated according to the first vital-sign feature. Whether the first feature signal is a first valid image signal is determined. When it is determined that the first feature signal is a first valid image signal, the first sub region is identified as a first ROI. In the frames occurring after the first frame, the first ROI is tracked.