Pulse rate extraction based on local-area motion magnification

Abstract Accurate pulse rate extraction based on complex wavelet transform (CWT) and local area magnification is proposed. The input video is passed through a wavelet denoising filter to remove/minimize the unwanted artifacts. A series of morphological operations are performed on the filtered output for extraction of local areas (which need magnification). The local area(s) are decomposed using CWT and Eulerian spatio-temporal processing is performed to magnify the invisible motion. The magnified local areas are combined with the filtered output and pulse rate is extracted based on Fourier analysis. Comparison with existing technique along-with pulse rate error analysis (with available ground truth) are performed to verify the proposed technique significance.