论文:2017,Vol:35,Issue(1):154-159
引用本文:
肖照林, 周果清. 场景深度无关的虚拟孔径图像鬼影去除算法[J]. 西北工业大学学报
Xiao Zhaolin, Zhou Guoqing. Depth Free Ghost Artifacts Reduction on Synthetic Aperture Imaging[J]. Northwestern polytechnical university

场景深度无关的虚拟孔径图像鬼影去除算法
肖照林1, 周果清2
1. 西安理工大学 计算机科学与工程学院, 陕西 西安 710048;
2. 西北工业大学 计算机学院, 陕西 西安 710072
摘要:
与传统成像系统相比,虚拟孔径成像的深度分辨率大为提高,但虚拟孔径图像的散焦区域普遍存在较严重的鬼影问题。提出一种基于熵值计算的鬼影去除算法,特点在于无须进行场景深度估计。首先,依据光场理论分析了鬼影产生原因;随后,提出了预测图像点聚焦趋势的假设,并依此逐像素计算熵值估计低通滤波器的核函数;根据相邻像素点相似性和相邻聚焦深度图像连续性,采用在空域低通滤波的方法实现图像鬼影去除。在仿真数据和真实数据的实验表明,该算法能够有效去除鬼影,并能保持虚拟孔径图像聚焦区域的清晰成像,从而验证了文中所提出算法的有效性。
关键词:    虚拟孔径成像    图像鬼影    熵值    计算成像   
Depth Free Ghost Artifacts Reduction on Synthetic Aperture Imaging
Xiao Zhaolin1, Zhou Guoqing2
1. School of Computer Science, Xi'an University of Technology, Xi'an 710048, China;
2. School of Computer Science, Northwestern Polytechnical University, Xi'an 710072, China
Abstract:
Compared to traditional imaging system, the depth resolution is significantly improved in the synthetic aperture photography (SAP) system. However, the ghost artifact is brought into defocused regions of SAP images, and that is an unexpected side effect in most applications. In this paper, we propose an entropy based ghost removal algorithm on SAP images. Our algorithm is free of depth estimation, which is still an open problem in computer vision. We first characterize the essence of ghost artifact by light field theory. Then, we detect the focus trend by introducing entropy based assumptions. We finally employ a spatial domain low pass filter to reduce the unexpected ghosting, in which the assumptions are used for estimating the kernel function. Experimental results show that ghosting artifacts is greatly smoothed by applying the proposed algorithm. In the mean time, the sharp focused regions are preserved in the focused region.
Key words:    Synthetic aperture photography    image ghost    entropy    computational imaging   
收稿日期: 2016-04-05     修回日期:
DOI:
基金项目: 国家自然科学基金(61501370、61401359、61272287)、陕西省自然科学基础研究计划(2016JQ6069、2015JQ6209)与虚拟现实技术与系统国家重点实验室开放课题(BUAAVR-16KF-10)资助
通讯作者:     Email:
作者简介: 肖照林(1984-),西安理工大学讲师、博士,主要从事计算机视觉、计算摄影学等研究。
相关功能
PDF(2787KB) Free
打印本文
把本文推荐给朋友
作者相关文章
肖照林  在本刊中的所有文章
周果清  在本刊中的所有文章

参考文献:
[1] Levoy M, Hanrahan P. Light Field Rendering[C]//ACM Siggraph, 1996:31-42
[2] Vaish V. Synthetic Aperture Imaging Using Dense Camera Arrays[D]. Stanford University, CA, USA, 2007
[3] Gortler S, Grzeszczuk R, Szeliski R, Cohen F M. The Lumigraph[C]//ACM Siggraph, 1996:43-54
[4] Chai J X, Tong X, Chan S C, Shum H Y. Plenoptic Sampling[C]//ACM Siggraph, 2000:307-318
[5] Georgiev T, Lumsdaine A. Depth of Field in Plenoptic Cameras[C]//Proc of Eurographics, 2009
[6] Bishop T E, Favaro P. The Light Field Camera:Extended Depth of Field Aliasing, and Super-Resolution[J]. IEEE TPAMI, 2011, 34(5):972-986
[7] Stewart J, Yu J, Gortler S J, McMillan L. A New Reconstruction Filter for Undersampled Light Fields[C]//Proc of Eurographics Workshop on Rendering, 2003:150-156
[8] Ng R. Fourier Slice Photography[C]//ACM Siggraph, 2005:735-744
[9] Levin A, Freeman W T, Durand F. Understanding Camera Trade-Offs through a Bayesian Analysis of Light Field Projections[C]//Proc of ECCV, 2008:88-101
[10] Hecht E. Optics[M]. 4th ed. San Francisco, Addison Wesley Press, 2001
[11] Liang C K, Shih Y C, Chen H H. Light Field Analysis for Modeling Image Formation[J]. IEEE Trans on Image Processing, 2011, 20(2):446-460
[12] Lumsdaine A, Georgiev T. Full Resolution Lightfield Rendering[R]. Adobe Technical Report, 2008