摘 要：在散焦模糊图像中单个像素的模糊量由其景深确定，因此散焦模糊是空间变化的。传统的去模糊方法在解决这种空间变化的反卷积问题的时候，会在不同散焦程度的边界处产生明显的振铃效应。针对上述存在的问题，文章提出了基于深度图像先验的散焦图像去模糊算法。实验结果表明，与现有的散焦图像去模糊算法对比，该算法具有更好的去散焦模糊的性能。
关键词：散焦图像；去模糊；振铃效应；深度图像先验
中图分类号：TP391 文献标识码：A 文章编号：2096-4706（2020）24-0084-05
Defocused Image Deblurring Based on Depth Image Prior
CHEN Tianming
（College of Electronic and Information Engineering，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，China）
Abstract：In a defocused and blurred image，the blur amount of a single pixel is determined by its depth of field，so the defocus blur is spatially variable. When traditional deblurring methods are used to solve the deconvolution problem of this kind of spatial variation，obvious ringing effect will appear at the boundary of different defocus degrees. Aiming at the above problems，this paper proposes a defocused image deblurring algorithm based on depth image prior. Experimental results show，comparing with the existing defocused image deblurring algorithm，this algorithm has better defocusing and deblurring performance.
Keywords：defocused image；deblurring；ringing effect；depth image prior
参考文献：
[1] LEMPITSKY V，VEDALDI A，ULYANOV D. Deep ImagePrior [C]//2018 IEEE/CVF Conference on Computer Vision and PatternRecognition.Salt Lake City：IEEE，2018：9446-9454.
[2] XU G D，QUAN Y H，JI H. Estimating Defocus Blur viaRank of Local Patches [C]//2017 IEEE International Conference onComputer Vision（ICCV）.Venice：IEEE，2017：5381-5389.
[3] KRISHNAN D，FERGUS R. Fast Image Deconvolutionusing Hyper-Laplacian Priors [C]//NIPS’09：Proceedings of the 22ndInternational Conference on Neural InformationProcessing Systems.Vancouver：Curran Associates Inc，2009：1033-1041.
[4] D’ANDRÈS L，SALVADOR J，KOCHALE A，et al. NonparametricBlur Map Regression for Depth of Field Extension [J].IEEETransactions on Image Processing，2016，25（4）：1660-1673.
[5] ZHUO S J，SIM T P. Defocus map estimation from a singleimage [J].Pattern Recognition，2011，44（9）：1852-1858.
[6] Tang C，HOU C P，Song Z J. Defocus map estimation from asingle image via spectrum contrast [J].Optics Letters，2013，38（10）：1706-1708.
[7] ZHANG X X，WANG R G，JING X B，et al. Spatiallyvariant defocus blur map estimation and deblurring from a single image [J].Journal of Visual Communication and ImageRepresentation，2016，35（3）：257-264.
[8] KARAALI A，JUNG C R. Edge-Based Defocus BlurEstimation With Adaptive Scale Selection [J].IEEE Transactions onImage Processing，2018，27（3）：1126-1137.
作者简介：陈天明（1995—），男，汉族，江苏启东人，硕士研究生，研究方向：数字图像处理。