基于改进K-means算法的彩色超声图像分割

Improved K-means Algorithm for the Color Ultrasound Image Segmentation

医学中的彩色超声图像受成像机理的影响,会出现对比度不高、边缘不清晰的现象。传统的图像分割方法存在处理结果准确率低、部分目标丢失等问题。文章针对图像分割中广泛应用的K-means算法依赖初始聚类中心和搜索易收敛于局部最优等不足,在基本粒子群算法中加入惯性权重来提高收敛性能,并用该算法确定初始聚类中心,解决K-means的缺陷,然后将改进算法应用于L*a*b颜色空间的彩色超声心脏图像分割中。实验显示,改进方法改善了聚类结果的准确率和稳定性,且聚类时间也短,对色彩度低的超声图像可取得很好的分割效果。

The color ultrasound image in medicine will appear low contrast and obscure edges phenomenon affected by imaging mechanism. The traditional image segmentation methods exist the problems of low accuracy,target missing partly and so on. For the lack of K - means algorithm widely used in image segmentation depending on the initial cluster centers and the search is easy to converge to a lo- cal optimum and so an,the inertia weight is added to the basic PSO to improve the convergence performance,and the algorithm is used to determine the initial cluster centers to solve the defect of K - means, then the improved algorithm is applied to the color ultrasound heart image segmentation of L ＊ a ＊ b color space. Experiment shows that the improved method improves the accuracy and stability of clustering results, and the clustering time is also low, it can achieve good segmentation results for the low color ultrasound images.