MRI图像中颅骨不连续外轮廓的多分辨率提取

A Multi-resolution Approach for Discontinuous Skull Contour Extraction from CT

目前 ,许多医学图像处理 (例如图像三维重构 ,图像建模 ,图像配准和图像融合等 )都需要进行图像的边缘提取 ,因此边缘提取方法在医学图像处理中有极其重要的意义。轮廓闭合图像的边缘比较容易提取 ,而解剖和外伤等原因造成轮廓不连续时边缘提取比较复杂。美国和欧洲等发达国家从 2 0世纪 90年代后期开始着手医学图像不连续边缘提取方法的研究 ,国内的有关研究仍少见报道。本课题以脑图像为例 ,试图综合运用多分辨率法与八邻距距离转换法来提取脑图像的不连续外轮廓。研究结果表明 ,此方法在提取不连续脑图像的边缘时快速有效 ,同时可以去掉高频干扰 ,有利于图像的配准。

At present, many methods on medical image processing (such as 3-D image reconstruction, image modeling, image registration and fusion, etc) need extraction of image contour. Therefore, it is important to study algorithm on contour extraction. Image contours could be classified into two types: closed ones and unclosed ones, depending upon anatomical structure or history of injury. It is much more complicated to extract a discontinuous contour than a closed one. In late 1990's, study on extracting discontinuous contour of medical images was started in USA and some European countries, while it is rarely reported in Asia. It was tried to utilize multi-resolutions and 8-neighbor DT to extract unconnected outer contour from CT images of brain, and satisfactory results are presented here.