An automatic 3D surface extraction in ultrasonic images

Presents an automatic 3D surface extraction method for ultrasonic images. In this technique, an image sequence is obtained along one axis of the 3D grid data. Based on the space continuity of the object, the region of interest (ROI) in each frame is determined by using a novel Polygon-based Tracking (PT) method. The edge in the ROI is then automatically extracted by the region growing technique, whose character vector is linearly constructed by the pixel value and the local entropy. Finally, the 3D surface is formed by recombination of the edges in a 2D image sequence. Experiments on water bubble ultrasonic images show that the new method is successful in detection of 3D surfaces.