融合全局与局部区域亮度的逆光图像增强算法  

Backlight Image Enhancement by Fusing Global and Local Region Brightness

作　　者：郭倩 朱振峰[1,2] 常冬霞 赵耀[1,2] 

机构地区：[1]北京交通大学信息科学研究所,北京100044 [2]北京市现代信息科学与网络技术重点实验室,北京100044

出　　处：《信号处理》2018年第2期140-147,共8页Journal of Signal Processing

基　　金：国家自然科学基金(61572068;61532005);教育部新世纪优秀人才支持计划项目(NCET-13-0661);中央高校基本科研业务费专项资金(2015JBM039)

摘　　要：逆光是造成图像质量降低的主要原因之一。针对逆光拍摄造成的亮度降低与细节信息损失等问题,本文提出了一种融合全局与局部区域亮度的逆光图像增强算法。通过颜色估计模型(CEM)对逆光图像进行全局增强,恢复图像细节信息。在此基础上,为避免颜色估计模型对逆光图像的局部区域进行增强时出现的"虚化"问题,建立了局部亮度保持的颜色估计模型。此外,为平衡全局与局部区域的增强性能,提出了一种基于图像局部块信息熵的自适应融合方法。实验结果验证了本文算法的有效性。Backlight is one of the main causes of degrading image quality, which limits its applications to a large extent. To address the issues of brightness reduction and loss of detail information of baeklight image, this paper proposes an image en- hancement algorithm by fusing global and local brightness enhancements. Through the color estimation model （CEM） based on global statistics, the detail information of backlight image can he well restored. To eliminate the fuzzy effect on dark re- gion of backlight image by CEM, the brightness preserving color estimation model （BPCEM） is presented. Unlike CEM, it is also shown that the problem of over enhancement on local region can be avoided by BPCEM. By taking advantages of both CEM and BPCEM model, an adaptive fusion method based on block information entropy is proposed. Thus, good balance between CEM and BPCEM can be made. Experimental results show that the proposed method has achieved competitive per- formance.

关 键 词：图像增强 颜色估计模型 图像融合 信息熵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]