东北大学学报:自然科学版 ›› 2020, Vol. 41 ›› Issue (9): 1231-1238.DOI: 10.12068/j.issn.1005-3026.2020.09.003

• 信息与控制 • 上一篇    下一篇

基于清晰度评价的自适应阈值图像分割法

张田, 田勇, 王子, 王昭东   

  1. (东北大学 轧制技术及连轧自动化国家重点实验室, 辽宁 沈阳110819)
  • 收稿日期:2019-12-10 修回日期:2019-12-10 出版日期:2020-09-15 发布日期:2020-09-15
  • 通讯作者: 张田
  • 作者简介:张田(1989-),男,安徽马鞍山人,东北大学博士后研究人员; 王昭东(1968-),男,安徽安庆人,东北大学教授,博士生导师.
  • 基金资助:
    国家重点研发计划项目(2018YFB1701600); 中央高校基本科研业务费专项资金资助项目(N170703010).

Adaptive Threshold Image Segmentation Based on Definition Evaluation

ZHANG Tian, TIAN Yong, WANG Zi, WANG Zhao-dong   

  1. State Key Laboratory of Rolling and Automation,Northeastern University,Shenyang 110819, China.
  • Received:2019-12-10 Revised:2019-12-10 Online:2020-09-15 Published:2020-09-15
  • Contact: ZHANG Tian
  • About author:-
  • Supported by:
    -

摘要: 阈值法是一种被广泛使用的图像分割方法.本文从图像中信息的变化情况出发,提出一种基于图像清晰度评价的新颖的自适应阈值分割方法.该方法采用清晰度评价函数作为阈值化后图像内灰度相似性变化的度量方法,通过反复迭代并结合皮尔逊相关性直至找到最佳的分割阈值.通过多组图像数据尤其低对比度图像,包括钢板表面轻微缺陷等图像进行了测试对比.结果表明:相比传统阈值分割方法及其改进算法,在低对比度图像的处理上,本文方法能够自适应地准确找到合理阈值,具有优异的图像分割性能.

关键词: 计算机视觉, 图像分割, 自适应阈值, 清晰度评价, 低对比度图像

Abstract: Threshold is a widely used method for image segmentation. With the variance of the information in the image, this paper proposed a novel adaptive threshold segmentation method based on image definition evaluation. This method uses the definition evaluation function as a measure of the gray similarity change in the image after thresholding. Repeated iteration and Pearson correlation were combined until the optimal segmentation threshold was found. Test comparisons were performed using multiple sets of image data, especially low-contrast images, such as slight defects on the steel surface. The results showed that compared with the traditional threshold segmentation method and its improved algorithm, in the processing of low-contrast images, the proposed method can adaptively and accurately find a reasonable threshold value, and has an excellent performance of image segmentation.

Key words: computer vision, image segmentation, adaptive threshold, definition evaluation, low-contrast image

中图分类号: