An Automatic Method of Extracting Contours from Ultrasound Medical Images

摘要

Because of the low resolution and few gray-level for ultrasound medical images, we can hardly obtain satisfied contours from images. In this paper, an automatic method of extracting contours from ultrasound images is proposed, which is the combination of wavelet transform and template matching. First the edges of serial images are detected by a threshold selection method based on discrete dyadic wavelet transform, that is we use wavelet transform on the histogram of images, the gray-level at the zero-crossing point after wavelet transform can be treated as threshold to detect edges of images. Second, we extract contours from serial images by a new contour template. According to the smooth transition characteristic of the neighboring images, the new contour template is obtained from current image, which is used to extract contour from the neighboring images. So all contours can be extracted from serial images quickly and automatically by this method. Experimental results indicates that the method proposed is an effective approach to gain accurate contours and is valuable for 3D surface reconstruction of ultrasound medical images.