Time-domain vibration measurement using phase-based motion detection with relative filter response

Abstract Motion amplification and detection using video capturing is widely used nowadays as an alternative to the conventional vibration measurement and analysis techniques. Still, vibration extraction process is cumbersome and prone to errors due to the limited video accuracy and motion capturing noise. In this work, a more advanced and efficient vibration measurement technique is presented. The response of the steerable pyramid filter of each video frame is divided by the response of the initial frame to obtain the low-pass modulation, which can be efficiently interpolated and differentiated to obtain best estimation of the phase gradient. The phase gradients of multiple points in a small window around each pixel and across different subbands are combined through weighted least squares optimization to obtain the motion. In addition, using comparative filter response eliminates the phase unwrapping process required to calculate the temporal phase accumulation. The simulation and experimental investigations have shown that the proposed algorithms can efficiently be used to estimate the motion.