Lagrangian Motion Magnification with Landmark-Prior and Sparse PCA for Facial Microexpressions and Micromovements

Video motion magnification methods are motion visualization techniques that aim to magnify subtle and imper-ceptibly small motions in videos. They fall into two main groups where Eulerian methods work on the pixel grid with implicit motion information and Lagrangian methods use explicitly estimated motion and modify point trajectories. The motion in high framerate videos of faces can contain a wide variety of information that ranges from microexpressions over pulse or respiratory rate to cues on speech and affective state. In his work, we propose a novel strategy for Lagrangian motion magnification that integrates landmark information from the face as well as an approach to decompose facial motions in an unsupervised manner using sparse PCA. We decompose the estimated displacements into different movement components that are subsequently amplified selectively. We propose two approaches: A landmark-based decomposition into global and local movements and a decomposition into multiple coherent motion components based on sparse PCA. Optical flow estimation is performed using a state-of-the-art deep learning-based method that we retrain on a microexpression database. Clinical relevance- This method could be applied to the annotation and analysis of micromovements for neurocognitive assessment and even novel, medical applications where micro-motions of the face might play a role.