动态场景的高动态范围图像生成  

High Dynamic Range Image Generation of Dynamic Scenes

作　　者：张登辉 霍永青 Zhang Denghui;Huo Yongqing(School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731)

机构地区：[1]电子科技大学信息与通信工程学院,成都611731

出　　处：《计算机辅助设计与图形学学报》2018年第9期1625-1636,共12页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61401072)

摘　　要：动态场景的多曝光高动态范围成像方法中,由于鬼影检测不准确导致鬼影去除失败,并且采用单帧参考图的鬼影去除方法生成的高动态范围图像存在细节信息不足.为此,提出一种无鬼影且高细节信息的多曝光高动态范围成像方法.首先通过相邻的两帧图像灰度值之间的顺序关系得到部分鬼影区域,另一部分鬼影区域则由相邻图像灰度值之间的差值,与经过多次迭代优化后的阈值判定得到,在分离相邻图像的鬼影区域后,整合为选定的单帧参考图像与其余输入图像之间的鬼影区域;然后采用图像修复方法增强去除移动物体后的非参考图像在鬼影区域内细节信息;最后使用静态场景的多曝光高动态范围成像方法生成高动态范围图像.实验结果表明,与已有方法相比,该方法不仅能有效地去除鬼影,同时避开了单帧参考图方法中鬼影区域细节信息完全由单帧参考图像决定的弊端,保留和改善了生成图像在鬼影区域内的细节信息.To solve the failure of ghost removal resulting from the inaccuracy of ghost detection and the de- tail insufficient in the generated high dynamic range image due to the single reference image-based ghost removal, a multi-exposure high dynamic range imaging method without ghost artifacts and high detail in- formation is proposed in this paper. Firstly, a partial ghost region is obtained by the order relationship be- tween the gray values of two adjacent frames, and the other ghost region is acquired by comparing the gray values of the difference image between two adjacent images with a threshold that is optimized by multiple iterations. After separating the ghost region of the adjacent images, the new ghost region between the se- lected single reference image and the rest of the input images is integrated by these images ghost region. Then, the image inpainting method is used to enhance the detail information of the non-reference image in the ghost region after removal of the moving objects; finally, a high dynamic range image is generated using a multi-exposure high dynamic range imaging method for static scenes. Experimental results show that compared with the existing methods, the proposed method not only can effectively remove ghost artifacts, but also avoids the drawback that the ghost region detail information in the single reference image method is completely determined by the single frame reference image, and preserves and improves the details of the generated image in the ghost region.

关 键 词：动态场景 多曝光高动态范围成像 鬼影检测 鬼影去除 高动态范围图像 图像修复 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]