电子学报 ›› 2022, Vol. 50 ›› Issue (1): 167-176.DOI: 10.12263/DZXB.20201040

• 学术论文 • 上一篇    下一篇

基于快速模糊聚类的动态多直方图均衡化算法

江巨浪, 刘国明, 朱柱, 黄忠, 郑江云   

  1. 安庆师范大学电子工程与智能制造学院, 安徽 安庆 246133
  • 收稿日期:2020-09-21 修回日期:2020-10-28 出版日期:2022-01-25 发布日期:2022-01-25
  • 作者简介:江巨浪 男,1967年生,安徽潜山人.现为安庆师范大学电子工程与智能制造学院教授,硕士生导师.主要研究方向为图像处理与分析、智能信息处理和机器学习.E-mail:jiangjulang@126.com
    刘国明 男,1989年生,安徽宣城人.现为安庆师范大学电子工程与智能制造学院硕士研究生.主要研究方向为图像处理与分析、机器学习.E-mail:807563263@qq.com
  • 基金资助:
    国家自然科学基金(61701006);安徽省自然科学基金(1708085QF147);安徽省自然科学基金(2108085MF196)

Dynamic Multi-Histogram Equalization Based on Fast Fuzzy Clustering

JIANG Ju-lang, LIU Guo-ming, ZHU Zhu, HUANG Zhong, ZHENG Jiang-yun   

  1. College of Electronic Engineering and Intelligent Manufacturing,Anqing Normal University,Anqing,Anhui 246133,China
  • Received:2020-09-21 Revised:2020-10-28 Online:2022-01-25 Published:2022-01-25

摘要:

为了提高直方图均衡化方法对不同亮度图像的适用性,提出一种基于图像聚类的动态多直方图均衡化算法.采用基于直方图加权的模糊C-均值聚类算法对图像进行快速聚类,并采用聚类质量评价指标确定最佳聚类个数.对于每个子图像的直方图,以像素数量均值作为幅度极值进行裁剪,根据原有灰度区间与像素占比重新分配动态范围.基于裁剪的直方图对每个子图像进行独立的均衡化,并映射到新的灰度区间.采用多种亮度特征的测试图像对算法性能进行了验证,并与直方图均衡化、递归均值分割的直方图均衡化、递归分割与加权的直方图均衡化、动态直方图均衡化、基于熵的动态子直方图均衡化、基于模糊分割的双直方图均衡化算法进行比较.实验结果表明,该算法增强的图像具有适中的对比度以及比其他算法更大的平均信息熵、更小的NIQE平均值.该算法对于多种亮度特征的图像均具有良好的适应性,能够有效避免过度增强与细节丢失,图像增强的视觉质量优于其他同类算法.

关键词: 图像增强, 直方图均衡化, 多直方图, 模糊聚类, 直方图裁剪, 动态范围分配

Abstract:

In order to improve the applicability of histogram equalization method to images of different brightness, a dynamic multi-histogram equalization algorithm based on image clustering is proposed. The fuzzy c-means clustering algorithm based on weighting of histogram is used for fast image clustering, and the optimal number of clusters is determined by clustering quality evaluation index. For the histogram of each sub image, clipping is done using the average number of pixels as the amplitude extremum, and the dynamic range is reallocated according to the original gray range and pixel proportion. Each sub-image is equalized individually based on the clipped histogram and mapped to a new gray range. The performance of the algorithm is verified by using test images with different luminance characteristics, and compared with other algorithms such as histogram equalization(HE), recursive mean separate histogram equalization(RMSHE), recursively separated and weighted histogram equalization(RSWHE), dynamic histogram equalization(DHE), entropy-based dynamic sub-histogram equalization(EDSHE), fuzzy-based histogram partitioning for bi-histogram equalization(FHPBHE). The experimental results show that the enhanced image by the proposed algorithm has moderate contrast, larger average information entropy and smaller average NIQE. The algorithm has good adaptability to various brightness features of the image, can effectively prevent over-enhancement and loss of details, and the visual quality of image enhancement is better than other similar algorithms.

Key words: image enhancement, histogram equalization(HE), multi-histogram, fuzzy clustering, histogram clipping, dynamic range allocation

中图分类号: