文献类型：期刊文献

英文题名：End-to-End Exposure Fusion Using Convolutional Neural Network

作者：Wang, Jinhua[1,2];Wang, Weiqiang[1];Xu, Guangmei[2];Liu, Hongzhe[3]

第一作者：王金华;Wang, Jinhua

通讯作者：Wang, JH[1];Wang, JH[2]

机构：[1]Univ Chinese Acad Sci, Sch Comp & Control Engn, Beijing 100049, Peoples R China;[2]Beijing Union Univ, Coll Informat Technol, Beijing 100101, Peoples R China;[3]Beijing Key Lab Informat Serv Engn, Beijing, Peoples R China

第一机构：Univ Chinese Acad Sci, Sch Comp & Control Engn, Beijing 100049, Peoples R China

通讯机构：[1]corresponding author), Univ Chinese Acad Sci, Sch Comp & Control Engn, Beijing 100049, Peoples R China;[2]corresponding author), Beijing Union Univ, Coll Informat Technol, Beijing 100101, Peoples R China.|[1141734]北京联合大学智慧城市学院;[11417]北京联合大学;

年份：2018

卷号：E101D

期号：2

起止页码：560-563

外文期刊名：IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS

收录：;EI(收录号：20180704785552);Scopus(收录号：2-s2.0-85041548195);WOS:【SCI-EXPANDED(收录号:WOS:000431762500034)】；

基金：This work was supported by National Nature Science Foundation of China (No. 91420202, No. 61572077, No. 61372148).

语种：英文

外文关键词：exposure fusion; convolutional neural networks; fusion rule; activity level measurement

摘要：In this paper, we describe the direct learning of an end-to-end mapping between under-/over-exposed images and well-exposed images. The mapping is represented as a deep convolutional neural network (CNN) that takes multiple-exposure images as input and outputs a high-quality image. Our CNN has a lightweight structure, yet gives state-of-the-art fusion quality. Furthermore, we know that for a given pixel, the influence of the surrounding pixels gradually increases as the distance decreases. If the only pixels considered are those in the convolution kernel neighborhood, the final result will be affected. To overcome this problem, the size of the convolution kernel is often increased. However, this also increases the complexity of the network (too many parameters) and the training time. In this paper, we present a method in which a number of sub-images of the source image are obtained using the same CNN model, providing more neighborhood information for the convolution operation. Experimental results demonstrate that the proposed method achieves better performance in terms of both objective evaluation and visual quality.