Applications of Phase-Based Motion Processing

Branch, Nicholas A ; Stewart, Eric C

关键词: OPTICAL FLOW (IMAGE ANALYSIS);    WAVELET ANALYSIS;    PIXELS;    SPATIAL FILTERING;    CAMERAS;    FOURIER TRANSFORMATION;    MAGNIFICATION;    NYQUIST FREQUENCIES;    PHASE SHIFT;    IMAGING TECHNIQUES;    BINOMIAL COEFFICIENTS;    SPHERICAL COORDINATES;    EIGENVALUES;   

RP-ID  :  M17-6418

美国|英语

来源: NASA Technical Reports Server

【 摘 要 】

Image pyramids provide useful information in determining structural response at low cost using commercially available cameras. The current effort applies previous work on the complex steerable pyramid to analyze and identify imperceptible linear motions in video. Instead of implicitly computing motion spectra through phase analysis of the complex steerable pyramid and magnifying the associated motions, instead present a visual technique and the necessary software to display the phase changes of high frequency signals within video. The present technique quickly identifies regions of largest motion within a video with a single phase visualization and without the artifacts of motion magnification, but requires use of the computationally intensive Fourier transform. While Riesz pyramids present an alternative to the computationally intensive complex steerable pyramid for motion magnification, the Riesz formulation contains significant noise, and motion magnification still presents large amounts of data that cannot be quickly assessed by the human eye. Thus, user-friendly software is presented for quickly identifying structural response through optical flow and phase visualization in both Python and MATLAB.