一种基于细节层分离的单曝光HDR图像生成算法  

High Dynamic Range Image Generating Algorithm Based on Detail Layer Separation of a Single Exposure Image

作　　者：张红英[1] 朱恩弘 吴亚东[2] ZHANG Hong-Ying;ZHU En-Hong;WU Ya-Dong(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621010;School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010)

机构地区：[1]西南科技大学信息工程学院,绵阳621010 [2]西南科技大学计算机科学与技术学院,绵阳621010

出　　处：《自动化学报》2019年第11期2159-2170,共12页Acta Automatica Sinica

基　　金：国家自然科学基金(61303127);四川省科技厅项目(2014SZ0223,2015GZ0212);西南科技大学龙山人才计划(17LZX445,17LZX426)资助~~

摘　　要：针对利用单幅低动态范围(Low dynamic range, LDR)图像生成高动态范围(High dynamic range, HDR)图像细节信息不足的问题,本文提出了一种基于细节层分离的单曝光HDR图像生成算法.该算法基于人类视觉系统模型,首先分别提取出LDR图像的亮度分量和色度分量,对伽马校正后的亮度分量进行双边滤波,提取出亮度分量的基本层,再对基本层和亮度分量进行遍历运算,得到亮度分量的细节层;然后,构造反色调映射函数,分别对细节层和伽马校正后的亮度图像进行扩展,得到各自的反色调映图像;之后,将反色调映射后亮度分量与压缩后的细节层进行融合,得到新的亮度分量.最后,融合色度分量与新的亮度分量,并对融合后图像进行去噪,得到最终的HDR图像.实验表明该算法能挖掘出部分隐藏的图像细节信息,处理效果较好,运行效率高,具有较好的鲁棒性.Aimed at the problem of insufficient information on high dynamic range(HDR) image generating using a single low dynamic range(LDR) image, an HDR image generating algorithm by means of detail layer separation of a single exposure image is proposed. Firstly, according to the human visual system model, the luminance component and chrominance component of the LDR image are extracted, respectively, then the gamma-corrected luminance component is filtered by bilateral filtering so as to extract the basic layer of the luminance component, and the extracted basic layer and the luminance component are traversed to get the detail layer of the luminance component. Secondly, the inverse tone mapping function is constructed to extend the detail image and the gamma-corrected luminance image to obtain the inverse tone mapping images, respectively. Thirdly, fusing the inverse tone mapping luminance component and the compressed detail layer obtains a new luminance component. Finally, the chromaticity component is combined with the new luminance component to get the fused image, which is de-noised to obtain the final HDR image. A comparison experiment shows that the proposed algorithm can excavate some hidden image detail information and has better processing effects, higher operation efficiency, and better robustness.

关 键 词：高动态范围 反色调映射算子 人类视觉系统 伽马校正 细节层分离 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]