Abstract
Many important physical phenomena involve subtle signals that are difficultto observe with the unaided eye, yet visualizing them can be very informative.Current motion magnification techniques can reveal these small temporalvariations in video, but require precise prior knowledge about the targetsignal, and cannot deal with interference motions at a similar frequency. Wepresent DeepMag an end-to-end deep neural video-processing framework based ongradient ascent that enables automated magnification of subtle color and motionsignals from a specific source, even in the presence of large motions ofvarious velocities. While the approach is generalizable, the advantages ofDeepMag are highlighted via the task of video-based physiologicalvisualization. Through systematic quantitative and qualitative evaluation ofthe approach on videos with different levels of head motion, we compare themagnification of pulse and respiration to existing state-of-the-art methods.Our method produces magnified videos with substantially fewer artifacts andblurring whilst magnifying the physiological changes by a similar degree.