Fast Wavelet based Ultrasound Image Segmentation for Liver Vessels Extraction

A method for extracting in real-time the position and size of liver vessels from 2D ultrasound images is described. It proceeds in two steps. Starting with a rough localization of the vessels using the scale-space theory, seed points are defined. Then, vessel contours are detected by analyzing the 1D wavelet transform of radial rays. The size of the vessel is then computed by fitting ellipses on the edge points.