Remote Heart Rate Estimation in Low-Light Environments Based on Eulerian Video Magnification

Remote measurement of heart rate based on videos is susceptible to light. In a low-light environment, conventional remote photoplethysmography (rPPG) algorithms (e.g., green, plane-orthogonal-to-skin (POS)) yield poor heart rate accuracy. An improved rPPG scheme is proposed by employing low-light image enhancement using the camera response model algorithm to enhance the frame of faces in low-light environments, and the Eulerian video magnification (EVM) to amplify the color change. The weak color variations induced by pulse on human skin surface can thus be accurately detected in low-light environment. In addition, it is found that after the EVM, using the green channel as the blood volume pulse signal yields the best overall results.