基于灰度幅值直方图的视网膜出血点检测算法
A Retinal Hemorrhages Detection Algorithm Based on Gray Amplitude Histogram
程雪琪 1苑玮琦1
作者信息
- 1. 沈阳工业大学视觉检测技术研究所,沈阳 110870;辽宁省机器视觉重点实验室,沈阳 110870
- 折叠
摘要
为减少图像亮暗不均、出血点间灰度差异大对视网膜出血点检测结果的影响,提出一种基于灰度幅值直方图零点阈值与区域特征分析的视网膜出血点检测算法.算法通过大尺度中值滤波器实现图像的阴影校正,结合凹凸线段对背景干扰点进行抑制,完成前期的预处理操作;由各行各列幅值直方图中第一个零点进行暗目标自动阈值分割,并基于幅值特征的区域生长改善血管连续性;根据血管与出血点间形状的不同去除血管,实现了出血点分割.在DIARETDB1公开图库中的测试结果验证了算法的有效性.
Abstract
In order to reduce the influence of uneven brightness and dark image and large gray difference between bleeding spots on the detection results of retinal bleeding spots,a retinal bleeding spot detection algorithm based on zero threshold of gray amplitude histogram and regional feature analysis is proposed.The algorithm realizes the shadow correction of the image through the large-scale median filter,and suppresses the background interference points by combining the concave-convex line segments to complete the pre-processing operation.Automatic threshold segmentation of dark targets is carried out by the first zero point in the amplitude histogram of each row and column,and the continuity of blood vessels is improved based on the region growth of amplitude characteristics.According to the different shapes between blood vessels and bleeding points,blood vessels are removed,and bleeding points are segmented.The test results in DIARETDB1 public gallery verify the effectiveness of the algorithm.
关键词
视网膜出血点/灰度幅值直方图/阈值分割/区域生长Key words
Retinal hemorrhages/Gray amplitude histogram/Threshold segmentation/Region-growing引用本文复制引用
出版年
2024
NETL
NSTL
万方数据