基于倒谱与频谱分析的模糊核估计算法

doi:10.16180/j.cnki.issn1007-7820.2019.02.002

摘要/Abstract

摘要：

图像盲去模糊是典型的图像和信号处理问题,其目的是从模糊图像中恢复出模糊核及清晰图像。在模糊核估计方面,以往的算法通常将模糊核尺度作为必要的输入参数,近年来有些算法虽然能较准确的估计参数化模糊核,但不能有效估计自然模糊图像中普遍存在的非参数化模糊核。文中利用图像梯度倒谱估计模糊核后再利用频谱分析以进一步精确的估计小尺寸模糊核的尺度。实验结果表明,文中提出的方法能适用于绝大多数场景下自然模糊图像的模糊核尺度估计。

关键词: 图像去模糊, 去卷积, 模糊核估计, 倒谱分析, 频谱分析, 傅里叶变换

Abstract:

Blind image deblurring is a classical image and signal processing problem, which aims to recover a blur kernel and a clear latent image from a blurry image. In the case of fuzzy kernel estimation, the size of blur kernel in earlier algorithms is an input parameter. In recent years, some algorithms can accurately estimate parameterized blur kernel. The none-parameterized blur kernel is more usual in the natural environment, while existing algorithms can not accurately deal with none-parameterized blur kernel. In this paper, cepstrum of image gradient was used to estimate the blur kernel and spectrum was used to further estimate the blur kernel with small size. Experiments results demonstrated that the proposed method could be applied to the fuzzy kernel-scale estimation of natural blurred images in most scenes.

Key words: image deblur, deconvolution, blur kernel estimation, cepstrum analysis, spectrum analysis, fourier transform

中图分类号:

TP391

王文恺. 基于倒谱与频谱分析的模糊核估计算法[J]. 电子科技, 2019, 32(2): 4-8.

WANG Wenkai. Blur Kernel Estimation Algorithm Based on Cepstrum and Spectrum Analysis[J]. Electronic Science and Technology, 2019, 32(2): 4-8.

图/表 5

图1

图2

图3

图4

表1

参考文献 17

[1]	Zhang J, Pan J, Lai W S , et al. Learning fully convolutional networks for iterative non-blind deconvolution[J].Arxiv, 2016(11):6969-6977.
[2]	Pan J, Hu Z, Su Z, et al. Deblurring text images via L0-regularized intensity and gradient prior [C]. Columbus:IEEE Conference on Computer Vision & Pattern Recognition, 2014.
[3]	Fang J, Yuan Y, Ji W, et al. Licence plate images deblurring with binarization threshold [C].Macau: IEEE International Conference on Imaging Systems & Techniques, 2015.
[4]	Liu H, Sun X, Fang L , et al. Deblurring saturated night image with function-form kernel[J]. IEEE Transactions on Image Processing, 2015,24(11):4637-4650. doi: 10.1109/TIP.2015.2461445 pmid: 26241971
[5]	Lu Q, Zhou W, Fang L , et al. Robust blur kernel estimation for license plate images from fast moving vehicles[J]. IEEE Transactions on Image Processing, 2016,25(5):2311-2323. doi: 10.1109/TIP.2016.2535375 pmid: 26955030
[6]	Xu L, Zheng S, Jia J. Unnatural L₀ sparse representation for natural image deblurring [C].Portland: IEEE Conference on Computer Vision & Pattern Recognition, 2013.
[7]	Xue F, Blu T . A novel SURE-based criterion for parametric PSF estimation[J]. IEEE Transactions on Image Processing, 2015,24(2):595-607. doi: 10.1109/TIP.2014.2380174 pmid: 25531950
[8]	Liu S, Wang H, Wang J , et al. Automatic blur-kernel-size estimation for motion deblurring[J]. Visual Computer International Journal of Computer Graphics, 2015,31(5):733-746. doi: 10.1007/s00371-014-0998-2
[9]	盛文正, 唐国红 . 基于匀速直线运动下模糊图像复原的研究[J]. 电子科技, 2014,27(12):58-60. doi: 10.3969/j.issn.1007-7820.2014.12.016
	Sheng Wenzheng, Tang Guohong . Blurred image restoration under uniform linear motion[J]. Electronic Science and Technology, 2014,27(12):58-60. doi: 10.3969/j.issn.1007-7820.2014.12.016
[10]	Lai W, Ding J, Lin Y, Chuang Y. Blur kernel estimation using normalized color-line priors [C].Santiago:IEEE International Conference on Computer Vision, 2015.
[11]	Yan R, Shao L . Blind image blur estimation via deep learning[J]. IEEE Transactions on Image Processing, 2016,25(4):1910-1921. doi: 10.1109/TIP.2016.2535273
[12]	Wu X, Xu H, Gui X, et al. Research on an algorithm of shape motion deblurring [C].Fuzhou: International Conference on Intelligent Computing, 2015.
[13]	Deshpande A, Patnaik S . Single image motion deblurring: An accurate PSF estimation and ringing reduction[J]. Optik-International Journal for Light and Electron Optics, 2014,125(14):3612-3618. doi: 10.1016/j.ijleo.2014.01.126
[14]	Lee J, Lee J, Park K , et al. Image deblurring based on the estimation of PSF parameters and the post-processing[J]. Optik-International Journal for Light and Electron Optics, 2013,124(15):2224-2228. doi: 10.1016/j.ijleo.2012.06.067
[15]	Deshpande A, Patnaik S. On improving accuracy of PSF estimation in spectral and cepstrum domain with morphological filtering [C].Surat:International Conference on Emerging Technology Trends in Electronics, Communication & Networking, 2012.
[16]	Shi M, Liu S . PSF estimation via gradient cepstrum analysis for image deblurring in hybrid sensor network[J]. International Journal of Distributed Sensor Networks, 2015,11(10):34-62. doi: 10.1155/2015/758034
[17]	Kohler L, Hirsch M, Mohler B, et al. Recording and playback of camera shake: benchmarking blind deconvolution with a real-world database [C].Florence: European Conference on Computer Vision, 2012.

模糊核	图片1	图片2	图片3	图片4	相对标准偏差
模糊核1	31×31	31×31	31×31	31×31	0.00%
模糊核2	31×31	31×31	31×31	31×31	0.00%
模糊核3	31×31	31×31	31×31	31×31	0.00%
模糊核4	31×31	31×31	31×31	31×31	0.00%
模糊核5	31×31	31×31	31×31	31×31	0.00%
模糊核6	31×31	31×31	51×51	31×31	59.81%
模糊核7	31×31	31×31	51×51	31×31	59.81%
模糊核8	201×109	201×101	183×87	157×87	21.27%
模糊核9	169×201	177×201	169×201	167×201	2.60%
模糊核10	145×103	201×133	161×97	147×71	40.92%
模糊核11	51×51	51×51	51×51	51×51	0.00%
模糊核12	31×31	51×51	51×51	31×31	53.16%