计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (13): 209-215.DOI: 10.3778/j.issn.1002-8331.1703-0117

• 图形图像处理 • 上一篇    下一篇

人眼灰度感知建模及其在图像增强中的应用

范晓鹏1,2,3,4,朱  枫1,3,4   

  1. 1.中国科学院 沈阳自动化研究所 光电信息技术研究室,沈阳 110016
    2.中国科学院大学,北京 100049
    3.中国科学院 光电信息处理重点实验室,沈阳 110016
    4.辽宁省图像理解与视觉计算重点实验室,沈阳 110016
  • 出版日期:2018-07-01 发布日期:2018-07-17

Human gray-scale perception modeling and its application in image enhancement

FAN Xiaopeng1,2,3,4, ZHU Feng1,3,4   

  1. 1.Optoelectronic Information Technology Laboratory, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China
    2.University of Chinese Academy of Sciences, Beijing 100049, China
    3.Key Laboratory of Opto-Electronics Information Processing, CAS, Shenyang 110016, China
    4.Key Lab of Image Understanding and Computer Vision, Liaoning Province, Shenyang 110016, China
  • Online:2018-07-01 Published:2018-07-17

摘要: 原直方图均衡化算法处理结果不理想,主要是图像灰度域上实现的均衡化并不代表人眼感知亮度域上的均衡化,对此提出人眼感知亮度域上的直方图均衡化算法,但由于使用的灰度人眼感知模型不准确,处理效果也没有得到明显的改善。提出一种灰度人眼感知模型的建模方法,利用实验测试的方法得到人眼视觉系统临界可见偏差曲线;再推导得到人眼对不同灰度背景下同等灰度差别的敏感度曲线;接着利用上一步结果再通过积分和归一化方法得到灰度人眼感知模型;根据以上灰度人眼感知模型对直方图均衡化算法进行改进。对比实验结果表明,提出算法相比于原直方图均衡化具有明显的改善效果,相比于CLAHE、BBHE以及HMF等直方图改进算法,也具有无需参数调节,增强效果显著和适应性强等优点。

关键词: 人眼视觉, 灰度感知, 直方图均衡化, 临界可见偏差, 图像增强

Abstract: The traditional histogram equalization algorithm is doing histogram equalization processing on the image gray domain, rather than the human vision perceived brightness domain, leading local image may be too bright or too dark. Some scholars have proposed histogram equalization improvement algorithms in the human eye perceived brightness field. However, due to the inaccurate use of the human gray-scale perception model, the test results have not been significantly improved. Based on this, a modeling method of gray eye model is proposed. Firstly, the just noticeable difference curve of the human visual system is obtained by the experimental test method. Secondly, the sensitivity curve of the human eye to the same gray scale difference in different gray backgrounds is obtained. Thirdly, gray-scale human perception model is obtained through integrating and normalizing method based on previous step’s result. Finally, it reimplements the improved histogram equalization algorithm according to the above human gray-scale perception model. The comparing experiments show that the proposed algorithm is significantly improved than the traditional histogram equalization, and has many advantages including no adjusting parameters, remarkable enhancement effect, strong adaptability compared with CLAHE, BBHE and HMF.

Key words: human vision, gray-scale perception, Histogram Equalization(HE), Just Noticeable Difference(JND), image enhancement