Study Of Filter And Segmentation In 3D Reconstruction Of Medical Ultrasound Image

3D reconstruction of medical ultrasound image is the main area in medical data 3D visualization, and an important research branch in science data visualization. It has been used more and more widely, because its characters of real time, lossless, cheapness, well repetition super sensitive.The author has collected and summarized the latest research achievement and development of three-dimensional reconstruction of medical images. In the thesis, we study some questions of the data pre-processing in 3D reconstruction of medical ultrasound image, such as acquiring of nephritic ultrasound image, Definition of ROI, filtering, segmentation, interpolation. According to the characteristic of speckle noise in ultrasound images, we present improved image edge preserving smooth filter and wavelet methods based on semi-soft threshold image de-noising. We make different processing in diverse resolution of image. The methods of semi-soft threshold embody the idea of multi-resolution together with adaptive process. Experiment result shows that the two methods can preserve more useful image edge and details to doctor as much as possible while removing noise.Across ultrasound image segmentation, we can receive apartment of anatomical proportion and get the position and figure of the focus. But image segmentation fall short of clinical request because of speckle and veins of apparatus of medical ultrasound image. Based on the particularity of ultrasound images sequence, this thesis imposed two methods to segment bladder ultrasound images. One method is based on region-growing and mathematical morphology algorithm, and the other method is based on radial edge and boundary tracking. The first method makes use of improved region-growing to avoid the orientation of holes which may be oriented by traditional region-growing algorithm. And then use mathematical morphology to adjust the segmented image. The second method make use of radial edge algorithm segment bladder ultrasound image to gained boundary character pixels .and then use boundary tracking algorithm to gained continuous boundary of bladder ultrasound images. Experiment result shows that two segmentation algorithms receive better results in segmentation of filtered ultrasound images. They are good at accuracy of segmentation and continuity of boundary.Finally, the correctness of the algorithms of data pre-processing in 3D reconstruction of ultrasound images is proved by such as bladder, cystic and embryo ultrasound images sequence. The results of experiment indicated that the algorithm and system of Pre-processing in 3D ultrasound image reconstruction from irregular sampled is feasible. Algorithms have definite theory meaning and applied value.