基于改进型八叉树分解的三维超声图像数据抽样方法

A data sampling method of 3-D ultrasound images based on improved octree decomposition

在医学图像处理中,由于三维超声图像数据具有海量、非均质的特点,使得处理过程复杂度增大,出现执行效率低等问题.因此,为实现海量数据处理的高效化和有效化,进行数据抽样是十分必要的.提出一种基于改进型八叉树分解的三维超声图像数据抽样方法,能自动高效地获得三维超声图像的高压缩率抽样数据.首先采用基于模糊集的灰度图像阀值分割算法确定分割阀值;然后,使用改进型八叉树算法对三维超声图像进行分解;最后根据选取准则输出最优同质立方体和典型异质立方体作为三维超声图像数据抽样结果.抽样方法充分考虑了抽样对象的空间关系,抽样结果代表性强且图像数据缩减到原始图像体积的1.758%,有效提高后续图像处理操作的运算效率.

In the field of medical image processing,3-D(three dimensional)ultrasound image data is characterized by large amount and heterogeneity,which makes the image data processing more complicated and time-consuming.Therefore,data sampling is necessary to achieve the high efficiency and fluency of mass data processing.This paper presents a data sampling method of 3-D ultrasound images based on improved octree decomposition,which can obtain representative data of 3-D ultrasound images automatically and efficiently.Firstly,the segmentation threshold can be determined by using image segmentation algorithm based on fuzzy set theory.Then the improved octree decomposition algorithm is applied to process 3-D ultrasound images.Finally,according to the selection criteria,an optimal homogeneous cube and a typical heterogeneous cube are selected as sampling results to output.This method takes fully consideration on the spatial relationship of sampling objects and shows an obvious practical effect as the sampling results has a strong representation and the image data was reduced to 1.758%of the original image volume,which can effectively improve the efficiency of the subsequent image processing.