﻿ 基于SVM的眼底血管分割技术
«上一篇
 文章快速检索 高级检索

 应用科技  2017, Vol. 44 Issue (3): 67-71  DOI: 10.11991/yykj.201605020 0

### 引用本文

LIANG Mingming, SHEN Liudi, WANG Yixue, et al. Eye fundus vessel segmentation technology based on SVM[J]. Applied Science and Technology, 2017, 44(3), 67-71. DOI: 10.11991/yykj.201605020.

### 文章历史

Eye fundus vessel segmentation technology based on SVM
LIANG Mingming, SHEN Liudi, WANG Yixue, ZHENG Liying
College of Computer Science and Technology, Harbin Engineering University, Harbin 150001, China
Abstract: An improved retinal vessel segmentation algorithm based on SVM was proposed to solve the problem that it is difficult to extract medium&small blood vessels at eye fundus.Firstly, the Gauss matched filter was used to filter retinal images and enhance the contrast ratio of images; then, in order to improve the efficiency of the filter, an improved method of Gaussian matched filter was put forward, which only needs the convolution of pixels and the optimum matching template; finally, in order to improve the classification performance of the algorithm, the mean drifting algorithm was used to preclassify the filtered image.The simulation results show that, by adopting the proposed method, the retinal vascular network can be segmented more accurately, especially, is is more accurate for the segment of medium&small blood vessels.
Key words: eye fundus images    blood vessel segmentation    Gaussian matched filter    linear structure detector    optimum matched template    SVM    mean drifting algorithm    image preclassification

1 改进的二维高斯匹配滤波 1.1 二维高斯匹配滤波

 $k\left( {x, y} \right) =-\exp \left( {-{x^2}/2{\sigma ^2}} \right)\left| y \right| \le L/2$ (1)

 ${\mathit{\boldsymbol{p}}_i} = \left[{u\;v} \right] = \left[{x\;y} \right] \cdot {\mathit{\boldsymbol{r}}_i} = \left[{x\;y} \right]\left[\begin{array}{l} \cos \;{\theta _i}\;\;-\sin \;{\theta _i}\\ \sin \;{\theta _i}\;\;\;\cos \;{\theta _i} \end{array} \right]$ (2)

 ${k_i}\left( {x, y} \right) =-\exp \left( {-{u^2}/2{\sigma ^2}} \right)\;\;\;\forall {\mathit{\boldsymbol{p}}_i} \in N$ (3)

 ${k_i}\left( {x, y} \right) = {k_i}\left( {x, y} \right)-{m_i}\;\;\forall {\mathit{\boldsymbol{p}}_i} \in N$ (4)

1.2 改进的二维高斯匹配滤波

 图 1 线检测器

1.3 实验结果与分析

 图 2 改进前后的图像滤波方法时间对比

2 改进的SVM眼底血管分割方法 2.1 高斯匹配滤波器增强图像

 图 3 滤波前后图像的对比

2.2 均值漂移算法预分类图像

 ${K_{{h_r}, {h_s}}} = \frac{C}{{h_s^2h_r^3}}k\left( {{{\left\| {\frac{{{x^s}}}{{{h_s}}}} \right\|}^2}} \right)k\left( {{{\left\| {\frac{{{x^r}}}{{{h_r}}}} \right\|}^2}} \right)$ (5)

 $\left\| {y_j^s-y_k^s} \right\| \le {h_s}$ (6)
 $\left\| {y_j^r-y_k^r} \right\| \le {h_r}$ (7)

 图 4 均值漂移算法在不同图片上的分类结果
2.3 基于SVM的血管分割

 图 5 文中方法流程
3 仿真结果与分析

 $\begin{array}{l} A = \left( {{T_{\rm{V}}} + {T_{\rm{B}}}} \right)/N\\ S = {T_{\rm{V}}}/\left( {{T_{\rm{V}}} + {F_{\rm{V}}}} \right)\\ C = {T_{\rm{B}}}/\left( {{T_{\rm{B}}} + {F_{\rm{B}}}} \right) \end{array}$

 图 6 改进前后方法在不同图像上的分割结果对比

4 结论

1) 对变换后的灰度图像做滤波增强处理，提高了中小血管的灰度级, 有利于准确提取特征。

2) 改进的滤波方法省去很多卷积运算和比较运算，使整个改进过程更加简单、快捷，尤其是处理海量数据图像时，能节省很多时间。

3) 利用均值漂移方法对滤波增强后的灰度图像做预分类处理，聚类后获得的区域个数少于原图像像素个数，采用SVM分类器对处理后的灰度图像进行分类收敛速度加快，节省了算法的运算时间，提高了分类性能。

 [1] MARTINEZ-PEREZ M E, HUGHES A D, STANTON A V, et al. Retinal vascular tree morphology:a semi-automatic quantification[J]. IEEE transactions on biomedical engineering, 2002, 49(8): 912-917. DOI:10.1109/TBME.2002.800789 (0) [2] KANSKI J J, TARRANT T R. Clinical ophthalmology:a systematic approach[J]. Archives of ophthalmology, 2003, 274(8): 1078-1085. (0) [3] KLONOFF D C, SCHWARTZ D M. An economic analysis of interventions for diabetes[J]. Diabetes care, 2000, 23(3): 390-404. DOI:10.2337/diacare.23.3.390 (0) [4] YAO Chang, CHEN Houjin. Automated retinal blood vessels segmentation based on simplified PCNN and fast 2D-Otsu algorithm[J]. Journal of central south university of technology, 2009, 16(4): 640-646. DOI:10.1007/s11771-009-0106-3 (0) [5] SOARES J V B, LEANDRO J J G, CESAR R M, et al. Retinal vessel segmentation using the 2-D Gabor wavelet and supervised classification[J]. IEEE transactions on medical imaging, 2006, 25(9): 1214-1222. DOI:10.1109/TMI.2006.879967 (0) [6] 贾广象, 陈莹. 基于图像目标特征空间自学习分类算法[J]. 计算机工程与应用, 2015, 51(20): 153-156. DOI:10.3778/j.issn.1002-8331.1310-0064 (0) [7] 朱承璋, 邹北骥, 向遥, 等. 彩色眼底图像视网膜血管分割方法研究进展[J]. 计算机辅助设计与图形学学报, 2015, 27(11): 2046-2057. (0) [8] 高旭. 眼底图像血管分割方法研究与实现[D]. 长沙: 中南大学, 2014. (0) [9] RICCI E, PERFETTI R. Retinal blood vessel segmentation using line operators and support vector classification[J]. IEEE transactions on medical imaging, 2007, 26(10): 1357-1365. DOI:10.1109/TMI.2007.898551 (0) [10] CHAUDHURI S, CHATTERJEE S, KATZ N, et al. Detection of blood vessels in retinal images using two-dimensional matched filters[J]. IEEE transactions on medical imaging, 1989, 8(3): 263-269. DOI:10.1109/42.34715 (0) [11] ZWIGGELAAR R, ASTLEY S M, BOGGIS C R M, et al. Linear structures in mammographic images:detection and classification[J]. IEEE transactions on medical imaging, 2004, 23(9): 1077-1086. DOI:10.1109/TMI.2004.828675 (0) [12] COMANICIU D, Meer P. Mean shift:a robust approach toward feature space analysis[J]. IEEE transactions on pattern analysis and machine intelligence, 2002, 24(5): 603-619. DOI:10.1109/34.1000236 (0) [13] CHENG Yizong. Mean shift, mode seeking, and clustering[J]. IEEE transactions on pattern analysis and machine intelligence, 1995, 17(8): 790-799. DOI:10.1109/34.400568 (0) [14] 许晓丽. 基于聚类分析的图像分割算法研究[D]. 哈尔滨: 哈尔滨工程大学, 2012. (0)