Computer Engineering and Applications ›› 2019, Vol. 55 ›› Issue (4): 193-199.DOI: 10.3778/j.issn.1002-8331.1711-0196

Previous Articles     Next Articles

Adaptive Multi-Exposure Image Fusion with Guided Filtering

XIE Wei1, WANG Liming1, HU Huanjun1, TU Zhigang2   

  1. 1.School of Computer, Central China Normal University, Wuhan 430079, China
    2.School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore
  • Online:2019-02-15 Published:2019-02-19


谢  伟1,王莉明1,胡欢君1,涂志刚2   

  1. 1.华中师范大学 计算机学院,武汉 430079
    2.南洋理工大学 电器电子工程学院,新加坡 639798

Abstract: In order to solve the phenomenon of halo and gradient inversion caused by guided filtering and the loss of edge of image fusion, this paper proposes a novel improved adaptive multi-exposure image fusion algorithm with guided filtering. Firstly, this paper sets the weight function according to the gradient information in the guided filtering, and combines the image pixel and the mean of the region to create the function, and realizes the adaption of the texture feature of different regions. Secondly, this paper sets the weight function by using the relationship between average brightness and contrast, saturation and exposure moderation, so that the weight value in the weighted average fusion process is no longer a fixed value, and it can adaptively adjust according to different image brightness, values are also different, making the fusion image quality better. Finally, the details of the original sequence map are superimposed on the improved guided filtering image, and the texture detail layer is constructed. The experimental results weaken the halo and gradient inversion phenomena, make the image more real and the details more clear, so the effect of image processing is better. The algorithm is superior to the multi-exposure fusion algorithm and the multi-exposure image fusion of the guided filter, which alleviates the halo phenomenon and obtains the highest 2.5%, 30% and 30% quality improvement respectively in the information entropy, mutual information and edge information evaluation.

Key words: halo, gradient inversion, average brightness, parameter self-adaption, detail enhancement

摘要: 针对引导滤波产生的光晕、梯度反转现象,以及图像融合边缘细节丢失的现象,提出一种改进引导滤波的自适应多曝光图像融合算法。在引导滤波中根据梯度信息设定权重函数,并结合图像像素点和一定区域的均值创建函数,共同实现不同区域的纹理特性自适应;利用平均亮度与对比度、饱和度及曝光适中度的关系,设置权值函数,使加权平均融合过程中的权重值不再是固定的数值,而能够根据不同的图像亮度自适应调整,权重值也不同,使得融合后的图像质量更好;将原序列图的细节信息叠加到改进的引导滤波图像中,构建纹理细节层。实验结果削弱了光晕及梯度反转现象,使图像更加真实,细节更加清晰,并且对有小光源的图像处理效果更好。算法结果明显优于多曝光融合算法及引导滤波的多曝光图像融合,在信息熵、互信息和边缘信息评价中分别取得最高2.5%、30%和30%左右的质量提升。

关键词: 光晕, 梯度反转, 平均亮度, 自适应参数调整, 细节增强