Multiscale approach for weighted least-squares optical flow for estimating arterial wall displacements

In this paper multiscale image decomposition was used to enhance the performance of weighted least-squares optical flow (WLSOF) in terms of estimating radial and longitudinal arterial wall displacements from B-mode ultrasound. For multiscale WLSOF (MWLSOF), ultrasound images were initially decomposed at one level using a 2D discrete wavelet transform, and WLSOF was applied on the resulting approximation images; the result was `translated´ to the original images through a coarse-to-fine transition process. WLSOF and MWLSOF were evaluated on synthetic image sequences of the common carotid artery. Multiscale image analysis increased the accuracy in displacement estimation, with MWLSOF yielding average displacement error reductions of 14% with respect to WLSOF. The methods were also effectively applied to real ultrasound image sequences of the carotid artery. It was concluded that MWLSOF can be efficiently used for estimating arterial wall displacements from B-mode ultrasound images.