电子学报 ›› 2017, Vol. 45 ›› Issue (5): 1165-1174.DOI: 10.3969/j.issn.0372-2112.2017.05.020

• 学术论文 • 上一篇    下一篇

图像梯度场双区间均衡化的细节增强

丁畅, 董丽丽, 许文海   

  1. 大连海事大学信息科学技术学院, 辽宁大连 116026
  • 收稿日期:2016-04-12 修回日期:2016-11-24 出版日期:2017-05-25
    • 通讯作者:
    • 董丽丽
    • 作者简介:
    • 丁畅 男,1988年11月出生于黑龙江哈尔滨.2014年于大连海事大学获得硕士学位.现为大连海事大学信息科学技术学院博士研究生,主要研究方向为图像处理、数据挖掘.E-mail:longlongclinton@163.com;许文海 男,1956年4月出生于吉林扶余.1991年于哈尔滨工业大学获得博士学位,1993年于东京工业大学获得博士学位.现为大连海事大学教授、博士生导师,主要研究方向为图像处理、光电信息与光电检测.E-mail:xuwenhai@dlmu.edu.cn
    • 基金资助:
    • 国家科技支撑计划 (No.2014BAB12B03); 国家自然科学基金 (No.61501077); 中央高校基本科研业务费专项资金 (No.3132016351)

Image Details Enhancement by Gradient Field Bi-interval Equalization

DING Chang, DONG Li-li, XU Wen-hai   

  1. School of Information Science and Technology, Dalian Maritime University, Dalian, Liaoning 116026, China
  • Received:2016-04-12 Revised:2016-11-24 Online:2017-05-25 Published:2017-05-25
    • Supported by:
    • National Key Technology Research and Development Program of the Ministry of Science and Technology (No.2014BAB12B03); National Natural Science Foundation of China (No.61501077); Fundamental Research Funds for the Central Universities (No.3132016351)

摘要:

直方图均衡化技术已广泛地应用于图像增强中,本文将该方法引入梯度域,这样可以使图像细节均匀地分布在各灰度级上.进一步本文又研究了图像梯度幅值分布特点,根据梯度幅值的右偏分布特性,采取双区间分别对梯度场进行均衡化处理.在区间阈值选取上,本文结合频率分布的数字特征,提出了区间阈值的两个选取准则并将梯度幅值合理地分为小梯度区间和大梯度区间.在目标梯度场复原过程中,本文提出矩阵变换法替代传统的"微分迭代法",减小了算法的时间复杂度.实验结果表明本文算法可使图像细节得到有效地增强.

关键词: 图像增强, 梯度场, 直方图均衡化, 双区间, 梯度幅值直方图, 偏度系数

Abstract:

The technique of histogram equalization (HE) is widely used in the field of image enhancement,the method is introduced into the gradient field in this paper,so image details are distributed in each gray level uniformly.Moreover,the characteristics of the histogram of gradient magnitude are studied,according to the right-skewed distribution of gradient magnitude,the gradient field is equalized by two intervals.In terms of the threshold's selection,numerical characteristics of frequency distribution are considered,two standards of the interval threshold are proposed,which can divide the gradient magnitude into the interval with low magnitude and the interval with high magnitude.On the process of the reconstruction of target gradient field,matrix transform method is proposed,which substitutes the traditional method "difference and iteration method",the proposed method can decrease the algorithm's time complexity.The experimental results show that the method this paper proposed can make image details enhance effectively.

Key words: image enhancement, gradient field, histogram equalization, bi-interval, the histogram of gradient magnitude, skewness

中图分类号: