Citation: | Wang Ronggui, Wang Qinghui, Yang Juan, et al. Image super-resolution reconstruction by fusing feature classification and independent dictionary training[J]. Opto-Electronic Engineering, 2018, 45(1): 170542. doi: 10.12086/oee.2018.170542 |
[1] | Wang L Z, Lu K, Liu P. Compressed sensing of a remote sensing image based on the priors of the reference image[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12(4): 736-740. doi: 10.1109/LGRS.2014.2360457 |
[2] | Wang Y H, Li J B, Fu P. Medical image super-resolution analysis with sparse representation[C]//Proceedings of the 2012 Eighth International Conference on Intelligent Information Hiding and Multimedia Signal Processing, 2012: 106-109. |
[3] | Li X, Orchard M T. New edge-directed interpolation[J]. IEEE Transactions on Image Processing, 2001, 10(10): 1521-1527. doi: 10.1109/83.951537 |
[4] | Zhang L, Wu X L. An edge-guided image interpolation algorithm via directional filtering and data fusion[J]. IEEE Transactions on Image Processing, 2006, 15(8): 2226-2238. doi: 10.1109/TIP.2006.877407 |
[5] | Dai S Y, Han M, Xu W, et al. SoftCuts: a soft edge smoothness prior for color image super-resolution[J]. IEEE Transactions on Image Processing, 2009, 18(5): 969-981. doi: 10.1109/TIP.2009.2012908 |
[6] | Sun J, Shum H Y. Image super-resolution using gradient profile prior: US 2010/0086227 A1[P]. 2010-04-08. |
[7] | Yang J C, Wright J, Huang T, et al. Image super-resolution as sparse representation of raw image patches[C]//Proceedings of the 2008 IEEE Conference on Computer Vision and Pattern Recognition, IEEE, 2008: 1-8. |
[8] | Yang J C, Wright J, Huang T S, et al. Image super-resolution via sparse representation[J]. IEEE Transactions on Image Processing, 2010, 19(11): 2861-2873. doi: 10.1109/TIP.2010.2050625 |
[9] | Yang J C, Wang Z W, Lin Z, et al. Coupled dictionary training for image super-resolution[J]. IEEE Transactions on Image Processing, 2012, 21(8): 3467-3478. doi: 10.1109/TIP.2012.2192127 |
[10] | Xu J, Qi C, Chang Z G. Coupled K-SVD dictionary training for super-resolution[C]//Proceedings of the 2014 IEEE International Conference on Image Processing, 2014: 3910-3914. |
[11] |
Zeyde R, Elad M, Protter M. On single image scale-up using sparse-representations[C]//Proceedings of the 7th International Conference on Curves and Surfaces, 2010: 711-730. |
[12] | Timofte R, De V, Van G L. Anchored neighborhood regression for fast example-based super-resolution[C]//Proceedings of the 2013 IEEE International Conference on Computer Vision, 2013: 1920-1927. |
[13] |
Timofte R, De S V, Van G L. A+: adjusted anchored neighborhood regression for fast super-resolution[C]//Proceedings of the 12th Asian Conference on Computer Vision, 2014: 111-126. |
[14] | Aharon M, Elad M, Bruckstein A. rmK-SVD: an algorithm for designing overcomplete dictionaries for sparse representation[J]. IEEE Transactions on Signal Processing, 2006, 54(11): 4311-4322. doi: 10.1109/TSP.2006.881199 |
[15] | Yang W M, Tian Y P, Zhou F, et al. Consistent coding scheme for single-image super-resolution via independent dictionaries[J]. IEEE Transactions on Multimedia, 2016, 18(3): 313-325. doi: 10.1109/TMM.2016.2515997 |
[16] | Yang S Y, Liu Z Z, Wang M, et al. Multitask dictionary learning and sparse representation based single-image super-resolution reconstruction[J]. Neurocomputing, 2011, 74(17): 3193-3203. doi: 10.1016/j.neucom.2011.04.014 |
[17] | Dong W S, Li X, Zhang L, et al. Sparsity-based image denoising via dictionary learning and structural clustering[C]//Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition, 2011: 457-464. |
[18] |
Feng X G, Milanfae P. Multiscale principal components analysis for image local orientation estimation[C]//Conference Record of the 36th Asilomar Conference on Signals, Systems and Computers, 2002, 1: 478-482. |
[19] |
Chang H, Yeung D Y, Xiong Y M. Super-resolution through neighbor embedding[C]//Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004: Ⅰ. |
[20] |
Bevilacqua M, Roumy A, Guillemot C, et al. Low-complexity single-image super-resolution based on nonnegative neighbor embedding[C]//Proceedings of British Machine Vision Conference (BMVC), 2012. |
[21] | Wang Z, 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): 600-612. doi: 10.1109/TIP.2003.819861 |
[22] | Sheikh H R, Bovik A C. Image information and visual quality[J]. IEEE Transactions on Image Processing, 2006, 15(2): 430-444. doi: 10.1109/TIP.2005.859378 |
[23] | Zhang L, Zhang L, Mou X Q, et al. FSIM: a feature similarity index for image quality assessment[J]. IEEE Transactions on Image Processing, 2011, 20(8): 2378-2386. doi: 10.1109/TIP.2011.2109730 |
Overview: Super-resolution reconstruction plays an important role in reconstructing image detail and improving image visual effects. It aims at reconstructing a high-resolution image from one or multiple low-resolution images. Because the high-resolution image contains more details than the low-resolution image, the high-resolution image possesses more values in remote sensing, medical diagnostic and so on. The super-resolution reconstruction can be divided into three categories: interpolation-based methods, reconstruction-based methods and learning-based methods. Because the learning-based methods can generate high-frequency details which are not available in the low-resolution, it attracts the attention of many researchers, and gradually becomes the main method of super-resolution reconstruction. We proposed a new effective super-resolution method based on the classification reconstruction and independent dictionary training. Most of the current classification of reconstruction methods are based on kmeans, but it tend to be difficult to further improve quality of reconstruction images due to the way of unsupervised clustering. So we analyze different image patches and three features including variance, and gradient feature, which are used to generate the decision tree for classification. Variance is used to distinguish smooth patches and other patches, gradient feature is used to distinguish stochastic patches and direction patches, and the last one is used to further distinguish directions. The experimental results show that our classification method has better classification results than kmeans. On the other hand, we train the high and low-resolution dictionaries based on K-SVD independently for different types of training sets. However, there is an obvious difference between the low-resolution coefficient and the high-resolution coefficient due to the independent dictionary training. Although the dictionary can represent respective training set well, the reconstructed images tend to be inferior if we insist on replacing the high-resolution coefficient using the low-resolution one. So it is necessary to solve their mapping function in order to achieve the better reconstruction effect. The mapping function is used to map the low-resolution coefficients to high-resolution coefficients during the reconstruction phase to ensure accurate and fast reconstruction of the image patches. We show the experimental results in detail when the magnification is 3, the experiment with different magnification and the comparative experiment with K-means. In general, the experimental results show that the proposed method has a significant improvement in the reconstruction effect compared with other classic methods.
The flowchart of super-resolution reconstruction
The decision tree of classification
Image set. (a) Hats; (b) Match; (c) Bird; (d) Foreman; (e) Sailboats; (f) Flowers; (g) Girl; (h) Butterfly; (i) Parrots