|
|
- 2018
基于暗通道先验和多方向加权TV的图像盲去模糊方法
|
Abstract:
针对全变差(TV)正则化图像复原其细节恢复能力有限且对噪声敏感等问题, 本文利用多方向边缘检测, 对传统TV模型进行改进, 得到基于边缘检测的多方向加权TV模型; 为了使复原模型更具普适性且提高细节恢复能力, 本文将暗通道先验融入上述模型, 提出基于暗通道先验和多方向加权TV的图像盲去模糊方法．同时, 在模糊核估计过程中, 提出了基于自适应强边缘提取的模糊核估计方法, 可有效剔除伪边缘、噪声等不利信息, 使模糊核估计更具鲁棒性; 最后, 给出了模糊核估计和去模糊模型的最优化求解算法．实验结果表明, 本文方法可准确估计模糊核, 复原图像含有更丰富的边缘、纹理等细节特征．
In order to overcome the limitations of traditional TV regularization in image restoration with the deficient ability of detail recovery and sensitivity to the noise，a novel multi-direction weighted TV(MDWTV)model is proposed based on multi-direction edge detection. Moreover，combing with the dark channel prior and MDWTV this paper presents a blind image deblurring method with more capacity of detail recovery and more applicability in various scenarios. At the same time，a kernel estimation method is proposed based on adaptive strong edge extraction which can remove fake edges and noise as well as increase the robustness of kernel estimation. Then a modified alternating direction method(ADM)is proposed to solve the above model. Extensive experimental results show that our method can estimate the blur kernel more accurately，and the restored image contains more details such as edge and texture
| [1] | Guo W, Yin W. EdgeCS:Edge guided compressive sensing reconstruction[J]. <i>Proceedings of SPIE</i>:<i>The International Society for Optical Engineering</i>, 2010, 7744:77440L-1-77440L-10. |
| [2] | Min L, Feng C. Compressive sensing reconstruction based on weighted directional total variation[J]. <i>Journal of Shanghai Jiaotong University</i>, 2017, 22(1):114-120. |
| [3] | Xu L, Zheng S, Jia J. Unnatural L0 sparse representation for natural image deblurring [C]// <i>Computer Vision and Pattern Recognition</i>. Portland, USA, 2013:1107-1114. |
| [4] | Pan J, Hu Z, Su Z, et al. L0-regularized intensity and gradient prior for deblurring text images and beyond[J]. <i>IEEE Transactions on Pattern Analysis & Machine Intelligence</i>, 2017, 39(2):342-355. |
| [5] | Wen Zaiwen, Yin Wotao, Zhang Hongchao, et al. On the convergence of an active-set method for ?1 minimization[J]. <i>Optimization Methods & Software</i>, 2012, 27(6):1127-1146. |
| [6] | Levin A, Weiss Y, Durand F, et al. Understanding and evaluating blind deconvolution algorithms[C]// <i>IEEE Conference on Computer Vision & Pattern Recognition</i>. Miami, USA, 2009:1964-1971. |
| [7] | Pan J, Liu R, Su Z, et al. Motion blur kernel estimation via salient edges and low rank prior[C]// <i>IEEE International Conference on Multimedia and Expo</i>. Chengdu, China, 2014:1-6.</i></i></i></i> |
| [8] | Babacan S D, Molina R, Do M N, et al. Bayesian blind deconvolution with general sparse image priors [C]// <i>European Conference on Computer Vision</i>. Florence, Italy, 2012:341-355. |
| [9] | Becerril J A, Rosenblueth J F. The importance of being normal, regular and proper in the calculus of variations [J]. <i>Journal of Optimization Theory & Applications</i>, 2017, 172(3):759-773. |
| [10] | Yang C X, Shao W Z, Huang L L. Boosting normalized sparsity regularization for blind image deconvolution[J]. <i>Signal Image & Video Processing</i>, 2017, 11(4):1-8. |
| [11] | Rudin L I, Osher S, Fatemi E. Nonlinear total variation based noise removal algorithms[J]. <i>Physica D Nonlinear Phenomena</i>, 1992, 60(1/2/3/4):259-268. |
| [12] | Kim H, Chen J, Wang A, et al. Non-local total-variation(NLTV)minimization combined with reweighted L1-norm for compressed sensing CT reconstruction[J]. <i>Physics in Medicine and Biology</i>, 2016, 61(18):6878-6891. |
| [13] | Chantas G, Galatsanos N P, Molina R, et al. Variational bayesian image restoration with a product of spatially weighted total variation image priors[J]. <i>IEEE Transactions on Image Processing</i>, 2010, 19(2):351-362. |
| [14] | Shan Q, Jia J, Agarwala A. High-quality motion deblurring from a single image[J]. <i>Acm Transactions on Graphics</i>, 2008, 27(3):15-19. |
| [15] | Candès E J, Wakin M B, Boyd S P. Enhancing sparsity by reweightedminimization[J]. <i>Journal of Fourier Analysis & Applications</i>, 2007, 14(5):877-905. |
| [16] | Chen Lixia, Song Guoxiang, Ding Xuanhao. Improved total variation algorithms to remove noise[J]. <i>Acta Photonica Sinica</i>, 2009, 38(4):1001-1004. |
| [17] | Cho S, Lee S. Fast motion deblurring[J]. <i>Acm Transactions on Graphics</i>, 2009, 28(5):1-8. |
| [18] | Fergus R, Singh B, Hertzmann A, et al. Removing camera shake from a single photograph[J]. <i>Acm Transactions on Graphics</i>, 2006, 25(25):787-794. |
| [19] | Ye X, Chen Y, Huang F. Computational acceleration for MR image reconstruction in partially parallel imaging [J]. <i>IEEE Transactions on Medical Imaging</i>, 2011, 30(5):1055-1063. |
| [20] | Michaeli T, Irani M. Blind deblurring using internal patch recurrence[C]// <i>European Conference on Computer Vision.<i> Zurich, Switzerland, 2014:783-798. |
| [21] | He K, Sun J, Tang X. Single image haze removal using dark channel prior[J]. <i>IEEE Transactions on Pattern Analysis & Machine Intelligence</i>, 2010, 33(12): |
| [22] | 2341-2353. |
| [23] | Pan J, Sun D, Pfister H, et al. Blind image deblurring using dark channel prior[C]// <i>IEEE Conference on Computer Vision and Pattern Recognition</i>. Las Vegas, USA, 2016:1628-1636. |
| [24] | Ng M K, Wang F, Yuan X. Inexact alternating driection methods for image recovery[J]. <i>Siam Journal on Scientific Computing</i>, 2011, 33(4):1643-1668. |
| [25] | Boykov Y, Kolmogorov V. Computing geodesics and minimal surfaces via graph cuts[C]// <i>IEEE International Conference on Computer Vision.<i> Nice, France, 2003:26-33. |
| [26] | Pan J, Liu R, Su Z, et al. Kernel estimation from salient structure for robust motion deblurring[J]. <i>Signal Processing Image Communication</i>, 2013, 28(9):1156-1170. |
