TR#400: Detecting and Segmenting Periodic Motion

Fang Liu and Rosalind W. Picard

A new algorithm is presented for the detection and segmentation of multiple periodic motions in image sequences. A periodicity measure which provides an accurate quantitative measure of signal periodicity in the form of temporal harmonic energy ratio is proposed. The periodicity templates which incorporate the segmented region with the associated periodicity energy distribution and motion fundamental frequencies provide a new means of periodic motion representation. The algorithm is computationally simple and robust in the presence of noise. Real world examples are used to illustrate the techniques.

Compressed Postscript . PDF . Full list of tech reports