Sparse sampling methods for efficient spatial coherence estimation

Abstract

Short-lag spatial coherence (SLSC) imaging, a coherence-based alternative beamforming technique, creates images related to the spatial covariance in backscatter. Because spatial covariance estimation is a computationally intensive process, efficient techniques are crucial to implementing SLSC imaging in real-time. Sparse sampling methods that take advantage of the statistical properties of spatial covariance of backscatter may yield adequate images at substantially faster rates. Among a variety of sparse spatial sampling techniques, a uniformly spaced sampling scheme was found to generate SLSC images with highest fidelity. Nearly identical images could be formed using a small fraction (17%) of the aperture, corresponding to a 97.3% reduction in the number of covariance computations. Furthermore, sparse temporal sampling was implemented by replacing the axial kernel of length T (used to compute covariance) with a single sample, and yielded SLSC images with improved axial resolution with a T -fold reduction in computation time. The two techniques were combined to obtain even further improvements in computation speed with little or no image degradation.