TwIST-TV regularization based image deblurring method
-
摘要: 传统的基于频域和小波域的去模糊算法所得的复原图像总是存在比较明显的边缘振铃及模糊效应，而较为有效的空域迭代优化去模糊算法速度通常比较慢。为了解决上述问题，提出了基于二步迭代阈值收缩（TwIST）与总变分（TV）约束相结合的图像去模糊算法（TwIST-TV）。首先在去模糊目标函数中加入对图像的TV 正则化约束，其次在对图像小波系数的每次二步迭代之前，加入对图像的TV 优化去噪约束，最后迭代获取去模糊图像。实验结果表明：相对于基于频域和小波域的模糊图像恢复算法，TwIST-TV 能有效抑制边缘模糊和振铃效应，复原图像的信噪比（SNR）、峰值信噪比（PSNR）高出1~7 dB，平均结构相似度指标（MSSIM）可高出0.05，相对于空域解卷积算法在保证求解精度相当的情况下具备6 倍以上的速度优势。Abstract: The deblurred images obtained by traditional frequency-wavelet domain based image restoration algorithms always result in prominent boundary ringing and smoothing artifacts. And the more effective space domain based alternating restoration algorithms usually work slowly. To overcome these problems, an algorithm named TwIST-TV which combines the two-step iterative shrinkage/thresholding (TwIST) and total variation (TV) regularization were proposed. This method first introduced the TV regularization constraint on the objective function, and then applied the TV-denoising method to regularize the midrestored image in each iteration before whose wavelet coefficients were processed by the TwIST method, and eventually obtained the deblurred image. Experimental results show that, in contrast to the frequencywavelet domain based image restoration algorithms, TwIST-TV can effectively suppress the boundary ringing and smoothing artifacts. The restored images can achieve 1-7 dB higher values of the signal-tonoise ratio (SNR), the peak signal-to-noise ratio (PSNR) and 0.05 higher value of the mean structural similarity (MSSIM) index. Proposed method has more than 6 times the speed advantage comparied with the methods which need alternating optimization in the space domain while maintain the accuracy of the solution.
-
Key words:
- image restoration /
- deblurring /
- wavelet transform /
- total variation /
- TwIST
-
[1] [2] Wu Zhenyu, Du Shaojun, Yao Hongli. Restoration and identification of defocus blurred image based on in-focus degree[J]. Infrared and Laser Engineering, 2011, 40 (4): 772-776. (in Chinese) [3] Le Xiang, Cheng Jian, Li Min. Improved approach to motion blur identification based on Radon transform[J]. Infrared and Laser Engineering, 2011, 40(5): 963-969. (in Chinese) [4] [5] [6] Wang Zhile, Zhao Ming, Li Bo, et al. Space variant image restoration based on maximum likelihood[J]. Infrared and Laser Engineering, 2012, 41(7): 1947-1951. (in Chinese) [7] [8] Levin A, Fergus R, Durand F, et al. Image and depth from a conventional camera with a coded aperture[J]. ACM Transactions on Graphics, 2007, 26(3): 70. [9] [10] Mallat S. A Wavelet Tour of Signal Processing[M]. New York: The Academic Press, 1998. [11] [12] Neelamani R, Choi H, Baraniuk R. ForWaRD: Fourier-wavelet regularized deconvolution for ill-conditioned systems[J]. IEEE Transactions on Signal Processing, 2004, 52(2): 418-433. [13] Meaarovic V Z, Galatsanos N P, Katsaggelos A K. Regularized constrained total least squares image restoration[J]. IEEE Transactions on Image Processing, 1995, 4 (8): 1096-1108. [14] [15] [16] Rudin L I, Osher S, Fatemi E. Nonlinear total variation based noise removal algorithms[J]. Physica D, 1992, 60: 259-268. [17] Chambolle A. An algorithm for total variation minimization and applications[J]. Mathematical Imaging and Vision, 2004, 20: 89-97. [18] [19] [20] Bioucas-dias J M, Figueiredo M A T. A new TwIST: two step iterative shrinkage/thresholding algorithms for image restoration[J]. IEEE Transactions on Image Processing, 2007, 16(12): 2992-3004. [21] [22] Wang Zhou, Bovik A C, Sheikh H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13 (4): 600-606. [23] Dong W D, Feng H J, Xu Z H, et al. Multi-frame blind deconvolution using sparse priors[J]. Optics Communications, 2012, 285: 2276-2288. -
点击查看大图
计量
- 文章访问数: 441
- HTML全文浏览量: 75
- PDF下载量: 214
- 被引次数: 0
下载:
