Robust motion estimation using complex principal components

This paper presents a novel motion estimator, termed the principal component displacement estimator (PCDE), which takes advantage of the signal separation capabilities of principal component analysis (PCA) to reject source signals representing decorrelation and noise. PCDE requires the computation of only a single principal component and operates on complex data, yielding computational speed in MATLAB on the same order or better than the commonly used Loupas algorithm. Synthetic ultrasound data were simulated to assess the performance of PCDE over a wide range of conditions. PCDE was also applied to experimental elastography data with reductions in the standard deviation of displacement estimates as large as 67% over other methods.