Image restoration approach based on structure dictionary learning
-
摘要: 本文提出一种新的结构字典学习方法，并利用它进行图像复原。首先给出结构字典学习的基本内容和方法，然后将傅里叶正则化方法和结构字典学习方法有效整合到图像复原算法中。结构字典学习方法是先将原图像进行结构分解，再分别学习出每个结构图像中的字典，最后利用这些字典对原图像进行稀疏的表示。结合傅里叶正则化，提出了一种有效的迭代图像复原算法：第一步在傅里叶域利用正则化反卷积方法得到图像的初步估计；第二步采用结构字典学习的方法对遗留的噪声进行去噪处理。实验结果表明，提出的方法在改进信噪比和视觉质量上都要优于6种先进的图像复原方法，改进的信噪比平均提升0.5 dB以上。Abstract: In this paper, we propose a new structure dictionary learning method, and perform image restoration based on this approach. First, we define the structure dictionary for the nature image. Second, an iterative algorithm is proposed with the decouple of deblurring and denoising steps in the restoration process, which effectively integrates the Fourier regularization and structure dictionary learning technique into the deconvolution framework. Specifically, we propose an iterative algorithm. In the deblurring step, we involve a regularized inversion of the blur in Fourier domain. Then we remove the remained noise using the structure dictionary learning method in the denoising step. Experiment results show that this approach outperforms 6 state-of-the-art image deconvolution methods in terms of improvement signal to noise rate (ISNR) and visual quality, and the ISNR can be improved by more than 0.5 dB.
-
Key words:
- structure dictionary /
- dictionary learning /
- image restoration /
- deconvolution
-
图 3 Exp3视觉效果对比图
Figure 3. Details of the image deconvolution experiment with a Lena in Exp3
图 4 Exp5视觉效果对比图
Figure 4. Details of the image deconvolution experiment with a Barbara in Exp5
图 5 Exp7实验效果对比图
Figure 5. Details of the image deconvolution experiment with a Barbara in Exp7
表 1 本文算法与现今主流算法ISNR值对比
Table 1. Output ISNR (dB) comparison of several state-of-the-art and the proposed deconvolution algorithms
Methods Exp1 Exp2 Exp3 Ex4 Exp5 Exp6 Exp7 ours 12.31 5.29 8.55 4.79 2.41 1.72 6.02 ForWaRD 9.56 3.85 6.97 3.50 0.94 0.98 4.02 FTVd 10.39 4.65 7.47 3.61 0.63 0.75 3.57 L0-ABS 11.06 4.80 7.79 4.22 0.73 0.81 3.98 SURE-LET 10.72 4.26 7.96 4.25 1.13 1.06 4.24 BSDL 6.99 4.88 4.83 4.50 2.00 1.13 2.65 BM3DDEB 10.85 4.56 7.97 4.37 1.90 1.28 5.86 
下载: 导出CSV
表 2 本文算法与现今主流算法实验运行时间 (秒) 对比
Table 2. Running time comparison of several state-of-the-art and the proposed deconvolution algorithm (second)
Methods 256×256 512×512 ours 31.20 130.64 ForWaRD 1.21 5.12 FTVd 0.63 3.25 L0-ABS 7.71 32.70 SURE-LET 0.33 1.43 BSDL 60.29 263.86 BM3DDEB 0.41 1.67
下载: 导出CSV
-
[1] 沈峘, 李舜酩, 毛建国, 等.数字图像复原技术综述[J].中国图像图形学报, 2009, 14(9):1764-1775. http://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200909012.htmSHEN H, LI S M, MAO J G, et al.. Digital image restoration techniques:a review[J]. J. Image and Graphics, 2009, 14(9):1764-1775.(in Chinese) http://www.cnki.com.cn/Article/CJFDTOTAL-ZGTB200909012.htm [2] 杨亚威, 胡双演, 张士杰, 等.基于字典对联合学习的退化图像复原方法[J].计算机辅助设计与图形学学报, 2015, 3:406-413. http://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201503005.htmYANG Y W, HU SH Y, ZHANG SH J, et al.. A degraded image restoration approach based on pairs of dictionaries jointly learning[J]. J. Computer-Aided Design & Computer Graics, 2015, 3:406-413.(in Chinese) http://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201503005.htm [3] 陈曦, 汪彦刚, 彭思龙, 等.部分模糊核已知的混合模糊图像复原算法[J].计算机辅助设计与图形学学报, 2010, 2:272-278. http://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201002015.htmCHEN X, WANG Y G, PENG S L, et al.. Restoration of degraded image from partially known mixed blur[J]. J. Computer-Aided Design & Computer Graphics, 2010, 2:272-2781.(in Chinese) http://www.cnki.com.cn/Article/CJFDTOTAL-JSJF201002015.htm [4] JAIN A K. Fundamental of Digital Image Processing[M]. Prentice-Hall. Inc., 1989:1420-1424. [5] 朱明, 杨航, 贺柏根, 等.联合梯度预测与导引滤波的图像运动模糊复原[J].中国光学, 2013, 6(6):850-855. http://www.chineseoptics.net.cn/CN/abstract/abstract9084.shtmlZHU M, YANG H, HE B G, et al.. Image motion blurring restoration of joint gradient prediction and guided filter[J]. Chinese Optics, 2013, 6(6):850-855.(in Chinese) http://www.chineseoptics.net.cn/CN/abstract/abstract9084.shtml [6] 刘成云, 常发亮.基于稀疏表示和Weber定律的运动图像盲复原[J].光学精密工程, 2015, 23(2):600-608. doi: 10.3788/OPE.LIU C Y, CHANG F L. Blind moving image restoration based on sparse representation and Weber's law[J]. Opt. Precision Eng., 2015, 23(2):600-608.(in Chinese) doi: 10.3788/OPE. [7] BANHAM M R, KATSAGGELOS A K. Digital image restoration[J]. IEEE Siggcal Processing Magazine, 1997, 14(2):24-41. doi: 10.1109/79.581363 [8] HANSEN P C. Rank-deficient and Discrete ill-posed Problems:Numerical Aspects of Linear Inversion[M]. Philadelphia, PA:SIAM, 1998. [9] WANG Y, YANG J, YIN W, et al.. Anew alternating minimization algorithm for total variation image reconstruction[J]. SIAM J. Imag. Sci., 2008, 1(3):248-272. doi: 10.1137/080724265 [10] OLIVEIRA J, BIOUCAS-DIAS J M, FIGUEIREDO M A, et al.. Adaptive total variation image deblurring:a majorization-minimization approach[J]. Signal Processing, 2009, 89(9):1683-1693. doi: 10.1016/j.sigpro.2009.03.018 [11] DAUBECHIES I, DEFRISE M, AND MOL C, et al.. An iterativethresholding algorithm for linear inverse problemswith a sparsity constraint[J]. Communications on Pure and Applied Mathematics, 2004, 57(11):1413-1457. doi: 10.1002/(ISSN)1097-0312 [12] BECK A, TEBOULLE M A. Fast iterativeshrinkage-thresholding algorithm for linear inverseproblems[J]. SIAM J. Imag. Sci., 2009(2):183-202. https://www.researchgate.net/publication/220124383_A_Fast_Iterative_Shrinkage-Thresholding_Algorithm_for_Linear_Inverse_Problems [13] BIOUCAS-DIAS J, FIGUEIREDO M. A new TwIST:two-step iterative shrinkage/thresholding algorithms for image restoration[J]. IEEE Trans. Image Process, 2007, 16(12):2992-3004. doi: 10.1109/TIP.2007.909319 [14] 张振东, 陈健, 王伟国, 等.基于SSIM_NCCDFT的超分辨率复原评价方法研究[J].液晶与显示, 2015, 30(4):713-721. doi: 10.3788/YJYXSZHANG ZH D, CHEN J, WANG W G, et al.. Evaluation method of super-resolution restoration based on SSIM_NCCDFT[J]. Chinese J. Liquid Crystals and Displays, 2015, 30(4):713-721.(in Chinese) doi: 10.3788/YJYXS [15] 郭萌, 赵岩, 王世刚, 等.基于区域选择的红外弱小目标超分辨率复原算法[J].液晶与显示, 2016, 31(4):415-420. doi: 10.3788/YJYXSGUO M, ZHAO Y, WANG SH G, et al.. Infrared dim-small target super-resolution restoration algorithm based on region selection[J]. Chinese J. Liquid Crystals and Displays, 2016, 31(4):415-420.(in Chinese) doi: 10.3788/YJYXS [16] NEELAMANI R, CHOI H, BARANIUK R G. ForWaRD:Fourier-wavelet regularized deconvolution for ill-conditioned systems[J]. IEEE Trans. Signal Process, 2004, 52(2):418-433.(in Chinese) doi: 10.1109/TSP.2003.821103 [17] PATEL V M, EASLEY G R, HEALY D M. Shearlet-based deconvolution[J]. IEEE Trans. Image Process, 2009, 18(12):2673-2685. doi: 10.1109/TIP.2009.2029594 [18] YANG H, ZHANG Z B. Fusion of wave atom-based wiener shrinkage filter and joint non-local means filter for texture-preserving image deconvolution[J]. Optical Engineering, 2012, 51(6):67-75. https://www.researchgate.net/publication/258687989_Fusion_of_wave_atom-based_Wiener_shrinkage_filter_and_joint_non-local_means_filter_for_texture-preserving_image_deconvolution [19] GUERRERO-COLON J A, MANCERA L, PORTILL J. Image restoration using space-variant Gaussian scale mixtures in overcomplete pyramids[J]. IEEE Trans. Image Process, 2007, 17(1):27-41. https://www.researchgate.net/publication/5621703_Image_Restoration_Using_Space-Variant_Gaussian_Scale_Mixtures_in_Overcomplete_Pyramids [20] FOI A, DABOV K, KATKOVNIK V, et al.. Shape-adaptive DCT for denoising and image reconstruction[J]. SPIE, 2006, 6064:203-214. https://www.researchgate.net/publication/240237642_Shape-adaptive_DCT_for_denoising_and_image_reconstruction_-_art_no_60640N [21] DABOV K, FOI A, KATKOVNIK V, et al.. Image denoising by sparse 3D transform-domain collaborative filtering[J]. IEEE Trans. Image Process, 2007, 16(8):2080-2095. doi: 10.1109/TIP.2007.901238 [22] DABOVE K, FOI A, Image restoration by sparse 3D transform-domain collaborative filtering[J]. SPIE, 2008, 6812:681207-12. [23] YANG H, ZHU M, ZHANG Z B, et al.. Guided filter based edge-preserving image non-blind deconvolution[C]. Proc. 20th IEEE International Conference on Image Processing (ICIP), Melbourne, Australia, 2013:4593-4596. [24] YANG H, ZHANG Z B, ZHU M, et al.. Edge-preserving image deconvolution with nonlocal domain transform″[J]. Optics and Laser Technology, 2013, 54:128-136. doi: 10.1016/j.optlastec.2013.05.020 [25] ELAD M, AHARON M. Image denoising via sparse and redundant representations over learned dictionaries[J]. IEEE Transactions on Image Processing, 2006, 15(12):3736-3745. doi: 10.1109/TIP.2006.881969 [26] COUZINIE-DEVY F, MAIRAL J, BACH F, et al.. Dictionary learning for deblurring and digital zoom[EB/OL].[2011-09-28].https://hal.inria.fr/inria-00627402. [27] JAIN A K, CAO K. Fingerprint image analysis:role of orientation patch and ridge structure dictionaries[J]. Geometry Driven Statistics, 2015:121-288. [28] LI K, SHAO F, JIANG G, et al.. Joint structure texture sparse coding for quality prediction of stereoscopic images[J]. Electronics Letters, 2015, 51(24):1994-1995. doi: 10.1049/el.2015.2049 [29] YUAN L, SUN J, QUAN L, et al.. Progressive inter-scale and intra-scale non-blind image deconvolution[J]. ACM Transaction on Graphics, 2008, 27(3):74-83. https://www.researchgate.net/publication/220183655_Progressive_inter-scale_and_intra-scale_non-blind_image_deconvolution [30] PORTILLA J. Image restoration through l0 analysis based sparse optimization in tight frames[C]. Proc. 16th IEEE ICIP, Cairo, Egypt, 2009:3909-3912. [31] NI J, TURAGA P, PATEL V M, CHELLAPPA R. Example-driven manifold priors for image deconvolution[J]. IEEE Trans. Image Process, 2011, 20(11):3086-3096. doi: 10.1109/TIP.2011.2145386 [32] XUE F, LUISIER F, BLU T. Multi-wiener SURE-LET deconvolution[J]. IEEE Trans. Image Process, 2013, 22(5):1954-1968. doi: 10.1109/TIP.2013.2240004 [33] CHATTERJEE P, MILANFAR P. Clustering-based denoising with locally learned dictionaries[J]. IEEE Trans. Image Process, 2009, 18(7):1438-1451. doi: 10.1109/TIP.2009.2018575 [34] MAIRAL J, BACH F, PONCE J, et al.. Non-local sparse models for image restoration[C]. Proc. IEEE Int. Conf. Comput. Vis., Kyoto, Japan, 2009:2272-2279. [35] LIU Q G, WANG S S, YING L L, et al.. Adaptive dictionary learning in sparse gradient domain for image recovery[J]. IEEE Transactions on Image Processing, 2013, 22(12):4652-4663. doi: 10.1109/TIP.2013.2277798 [36] STARCK J L, ELAD M, DONOHO D, et al.. Redundant multiscale transforms and their application for morphological component analysis[J]. J. Advances in Imaging and Electron Physics, 2004, 132:287-348. doi: 10.1016/S1076-5670(04)32006-9 [37] ZHANG Q, SHEN X Y, XU L, et al.. Rolling guidance filter[C]. Europ. Conf.Comput.Vision. Springer, 2014:815-830. [38] YANG H, ZHU M, WU X T, et al.. Dictionary learning approach for image deconvolution with variance estimation[J]. Applied Optics, 201453(24):5677-5683. https://www.researchgate.net/publication/265057447_Dictionary_learning_approach_for_image_deconvolution_with_variance_estimation -
下载:
图(5) / 表(2)
计量
- 文章访问数: 1780
- HTML全文浏览量: 422
- PDF下载量: 901
- 被引次数: 0
