首页 » 文章 » 文章详细信息
Mathematical Problems in Engineering Volume 2017 ,2017-11-27
A l1 Norm Based Image Prior Combination in Multiframe Superresolution
Research Article
Lei Min 1 , 2 , 3 , 4 Ping Yang 1 , 3 Lizhi Dong 1 , 3 Wenjin Liu 1 , 3 Shuai Wang 1 , 3 Bing Xu 1 , 3 Yong Liu 2
Show affiliations
DOI:10.1155/2017/2694638
Received 2017-07-08, accepted for publication 2017-10-11, Published 2017-10-11
PDF
摘要

We address the multiframe superresolution problem using the variational Bayesian method in this paper. In the variational Bayesian framework, the prior is crucial in transferring the ill-posed reconstruction problem to a well-posed one. We propose a prior combination method based on filter bank and l1 norm. Multiple filters are used in our prior model, and the corresponding combination coefficient vector can be estimated by the characteristics of the filtered image and noise. Furthermore, the local adaptive coefficients of every filter are more effective in removing noise and preserving image edges. Extensive experiments demonstrate the advantages of the proposed method.

授权许可

Copyright © 2017 Lei Min et al. 2017
This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

图表

The imaging process.

Images used in the synthetic experiments.

Images used in the synthetic experiments.

Images used in the synthetic experiments.

Images used in the synthetic experiments.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using first-order combinations: case 1 f1,f2, case 2 f14,f15,f16, case 3 f3,f4, case 4 f10,f11, and case 5 f12,f13.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using first-order combinations: case 1 f1,f2, case 2 f14,f15,f16, case 3 f3,f4, case 4 f10,f11, and case 5 f12,f13.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using second-order combinations: case 1 f6,f7, case 2 f5, and case 3 f8,f9.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using second-order combinations: case 1 f6,f7, case 2 f5, and case 3 f8,f9.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using first- and second-order combinations: case 1 f1,f2,f6,f7, case 2 f1,f2,f8,f9, case 3 f1,f2,f5, case 4 f14,f15,f16,f6,f7, case 5 f14,f15,f16,f8,f9, case 6 f14,f15,f16,f5, case 7 f3,f4,f6,f7, case 8 f3,f4,f8,f9, and case 9 f3,f4,f5.

(a) PSNR and (b) SSIM of the reconstructed images from the image in Figure 2(b), for our proposed method, using first- and second-order combinations: case 1 f1,f2,f6,f7, case 2 f1,f2,f8,f9, case 3 f1,f2,f5, case 4 f14,f15,f16,f6,f7, case 5 f14,f15,f16,f8,f9, case 6 f14,f15,f16,f5, case 7 f3,f4,f6,f7, case 8 f3,f4,f8,f9, and case 9 f3,f4,f5.

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

PSNR and SSIM for the reconstructed images for the images in Figure 2 with different method: (a)-(b) for Figure 2(a), (c)-(d) for Figure 2(b), (e)-(f) for Figure 2(c), and (g)-(h) for Figure 2(d).

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstructions on the EIA sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 5 frames, magnification factor 2). (a) The first four low resolution frames. Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

Superresolution reconstruction on the car sequence (L = 15 frames, magnification factor 2). (a) Four low resolution frames (frames 1, 4, 7, and 11). Reconstruction using superresolution method: (b) BBC, (c) SAR, (d) TV, (e) NS_NF3, (f) NF2, and (g) NF4.

通讯作者

Ping Yang.Key Laboratory on Adaptive Optics, Chinese Academy of Sciences, Chengdu 610209, China, cas.cn;The Institute of Optics and Electronics, The Chinese Academy of Sciences, Chengdu 610209, China, cas.cn.pingyang2516@163.com

推荐引用方式

Lei Min,Ping Yang,Lizhi Dong,Wenjin Liu,Shuai Wang,Bing Xu,Yong Liu. A l1 Norm Based Image Prior Combination in Multiframe Superresolution. Mathematical Problems in Engineering ,Vol.2017(2017)

您觉得这篇文章对您有帮助吗?
分享和收藏
11

是否收藏?

参考文献
[1] K. Malczewski, R. Stasinski. Toeplitz-based iterative image fusion scheme for MRI. :341-344. DOI: 10.1109/TGRS.2007.893271.
[2] H. Greenspan. (2009). Super-resolution in medical imaging. The Computer Journal.52(1):43-63. DOI: 10.1109/TGRS.2007.893271.
[3] S. Peleg, D. Keren, L. Schweitzer. (1987). Improving image resolution using subpixel motion. Pattern Recognition Letters.5(3):223-226. DOI: 10.1109/TGRS.2007.893271.
[4] R. M. Willett, I. Jermyn, R. D. Nowaketal. (2004). Wavelet-based super-resolution in astronomy. Astronomical Data Analysis Software & Systems XIII.314, article 107. DOI: 10.1109/TGRS.2007.893271.
[5] S. D. Babacan, R. Molina, A. . Katsaggelos. (2011). Variational Bayesian super resolution. IEEE Transactions on Image Processing.20(4):984-999. DOI: 10.1109/TGRS.2007.893271.
[6] X. Zhang, M. Tang, R. Tong. (2012). Robust super resolution of compressed video. The Visual Computer.28(12):1167-1180. DOI: 10.1109/TGRS.2007.893271.
[7] S. Villena, M. Vega, D. Babacan. . DOI: 10.1109/TGRS.2007.893271.
[8] C. Liu, D. Sun. (2014). On bayesian adaptive video super resolution. IEEE Transactions on Pattern Analysis and Machine Intelligence.36(2):346-360. DOI: 10.1109/TGRS.2007.893271.
[9] S. Villena, M. Vega, R. Molina, A. K. Katsaggelos. et al.(2014). A non-stationary image prior combination in super-resolution. Digital Signal Processing.32:1-10. DOI: 10.1109/TGRS.2007.893271.
[10] M. T. Merino, J. Núñez. (2007). Super-resolution of remotely sensed images with variable-pixel linear reconstruction. IEEE Transactions on Geoscience and Remote Sensing.45(5):1446-1457. DOI: 10.1109/TGRS.2007.893271.
[11] A. Tekalp, M. Ozkan, M. Sezan. High-resolution image reconstruction from lower-resolution image sequences and space-varying image restoration. :169-172. DOI: 10.1109/TGRS.2007.893271.
[12] M. Beal. (2003). Variational algorithms for approximate Bayesian inference [Ph.D. dissert]. DOI: 10.1109/TGRS.2007.893271.
[13] H. Shen, L. Peng, L. Yue, Q. Yuan. et al.(2016). Adaptive Norm Selection for Regularized Image Restoration and Super-Resolution. IEEE Transactions on Cybernetics.46(6):1388-1399. DOI: 10.1109/TGRS.2007.893271.
[14] S. Villena, M. Vega, R. Molina, A. K. Katsaggelos. et al.Image prior combination in super-resolution image reconstruction. :616-620. DOI: 10.1109/TGRS.2007.893271.
[15] M. Elad, A. Feuer. (1997). Restoration of a single superresolution image from several blurred, noisy, and undersampled measured images. IEEE Transactions on Image Processing.6(12):1646-1658. DOI: 10.1109/TGRS.2007.893271.
[16] P. Cheeseman, B. Kanefsky, R. Kraft, J. Stutz. et al.(1994). Super-resolved surface reconstruction from multiple images. Technical Report FIA9412. DOI: 10.1109/TGRS.2007.893271.
[17] L. Yue, H. Shen, J. Li, Q. Yuan. et al.(2016). Image super-resolution: The techniques, applications, and future. Signal Processing.128:389-408. DOI: 10.1109/TGRS.2007.893271.
[18] K. Nasrollahi, T. B. Moeslund. (2014). Super-resolution: a comprehensive survey. Machine Vision and Applications.25(6):1423-1468. DOI: 10.1109/TGRS.2007.893271.
[19] H. Shen, L. Zhang, B. Huang, P. Li. et al.(2007). A MAP approach to joint motion estimation, segmentation, and super resolution. IEEE Transactions on Image Processing.16(2):479-490. DOI: 10.1109/TGRS.2007.893271.
[20] L. Zhang, Q. Yuan, H. Shen, P. Li. et al.(2011). Multiframe image super-resolution adapted with local spatial information. Journal of the Optical Society of America A: Optics and Image Science, and Vision.28(3):381-390. DOI: 10.1109/TGRS.2007.893271.
[21] R. Y. Tsai, T. S. Huang. (1984). Multi-frame image restoration and registration. Advances in Computer Vision & Image Processing.1:317-339. DOI: 10.1109/TGRS.2007.893271.
[22] S. Farsiu, M. D. Robinson, M. Elad, P. Milanfar. et al.(2004). Fast and robust multiframe super resolution. IEEE Transactions on Image Processing.13(10):1327-1344. DOI: 10.1109/TGRS.2007.893271.
[23] S. Villena, M. Vega, R. Molina, A. Katsaggelos. et al.Bayesian Super-Resolution image reconstruction using an 1 prior. :152-157. DOI: 10.1109/TGRS.2007.893271.
[24] M. K. Ng, H. Shen, E. Y. Lam, L. Zhang. et al.(2007). A total variation regularization based super-resolution reconstruction algorithm for digital video. EURASIP Journal on Advances in Signal Processing.2007, article 74585. DOI: 10.1109/TGRS.2007.893271.
[25] B. Lucas, T. Kanade. An iterative image registration technique with an application to stereo vision. :121-130. DOI: 10.1109/TGRS.2007.893271.
[26] R. Molina, J. Núñez, F. J. Cortijo, J. Mateos. et al.(2001). Image restoration in astronomy: a bayesian perspective. IEEE Signal Processing Magazine.18(2):11-29. DOI: 10.1109/TGRS.2007.893271.
[27] P. Milanfar. MDSP Super-Resolution And Demosaicing Datasets. DOI: 10.1109/TGRS.2007.893271.
[28] J. Bioucas-Dias, M. Figueiredo, J. Oliveira. Total variation-based image deconvolution: a majorization-minimization approach. :II-861-II-864. DOI: 10.1109/TGRS.2007.893271.
文献评价指标
浏览 192次
下载全文 10次
评分次数 0次
用户评分 0.0分
分享 11次