基于稀疏贝叶斯估计的单图像超分辨率算法
|
摘要
针对现有超分辨率方法对不同低分辨率图像的超分辨率效果差异较大的问题,提出了一种基于稀疏贝叶斯估计的单图像超分辨率方法。该方法将单图像超分辨率问题看做是回归问题,采用Kronecker脉冲函数作为回归基函数,综合利用图像的局部信息和全局信息寻找特定预测的最优稀疏解决方案,采用贝叶斯方法估计权重,据此重构超分辨率图像。实验结果表明,采用该方法对14幅测试图像运行单图像超分辨率算法,得到的平均峰值信噪比高、方差小、耗时少,证实了该方法的超分辨率效果好、适应性强,且运算效率高。
Aiming at the problem that the super-resolution effect on different low-resolution images of existing super-resolution methods has large difference,this paper proposed a new single image super-resolution method based on sparse Bayesian estimation. In this method,it regarded the single image super-resolution problem as a regression problem. It used the Kronecker pulse functions as the regression basis functions,and obtained the optimal sparse solution of the specific prediction by combining the local information and global information of the image. It used the Bayesian method to estimate the weights,thereby reconstructing the super-resolution image. The experimental results show that this method can obtain high average peak signal to noise ratio,small variance and less time-consuming,when carried out on 14 testing images for single image super-resolution. It proves that this method has good super-resolution effect,strong adaptability,and high efficiency.
Aiming at the problem that the super-resolution effect on different low-resolution images of existing super-resolution methods has large difference,this paper proposed a new single image super-resolution method based on sparse Bayesian estimation. In this method,it regarded the single image super-resolution problem as a regression problem. It used the Kronecker pulse functions as the regression basis functions,and obtained the optimal sparse solution of the specific prediction by combining the local information and global information of the image. It used the Bayesian method to estimate the weights,thereby reconstructing the super-resolution image. The experimental results show that this method can obtain high average peak signal to noise ratio,small variance and less time-consuming,when carried out on 14 testing images for single image super-resolution. It proves that this method has good super-resolution effect,strong adaptability,and high efficiency.
引文
[1] Liu Ding,Wang Zhaowen,Wen Bihan,et al. Robust single image super-resolution via deep networks with sparse prior[J]. IEEE Trans on Image Processing,2016,25(7):3194-3201.
[2] Yang M C,Wang Y C F. A self-learning approach to single image superresolution[J]. IEEE Trans on Multimedia,2013,15(3):498-508.
[3] Wang Lingfeng,Xiang Shiming,Meng Gaofeng,et al. Edge-directed single-image super-resolution via adaptive gradient magnitude self-interpolation[J]. IEEE Trans on Circuits&Systems for Video Technology,2013,23(8):1289-1299.
[4] Zhang Kaibing,Gao Xinbo,Tao Dacheng,et al. Single image superresolution with multiscale similarity learning[J]. IEEE Trans on Neural Networks&Learning Systems,2013,24(10):1648-1659.
[5] Rueda A,Malpica N,Romero E. Single-image super-resolution of brain MR images using overcomplete dictionaries[J]. Medical Image Analysis,2013,17(1):113-132.
[6] Zhang Kaibing,Tao Dacheng,Gao Xinbo,et al. Learning multiple linear mappings for efficient single image super-resolution[J]. IEEE Trans on Image Processing,2015,24(3):846-860.
[7]李银辉,吕晓琪,于荷峰.基于l1和l2混合范式的序列图像超分辨率重建[J].计算机应用,2015,35(3):840-843.(Li Yinhui,Lyu Xiaoqi,Yu Hefeng. Sequence images super-resolution reconstruction based on l1and l2mixed norm[J]. Journal of Computer Applications,2015,35(3):840-843.)
[8]郭琳,叶波,陈庆虎,等.融合几何变换相似块的序列图像超分辨率重建[J].光电工程,2015,42(8):79-85.(Guo Lin,Ye Bo,Chen Qinghu,et al. Super-resolution reconstruction of image sequences via fusing similar patches with geometric transformation[J].Opto-Electronic Engineering,2015,42(8):79-85.)
[9]丁兰,傅志中,李晓峰,等.使用非均匀采样复原理论进行裂纹修复的混合分辨率多视角图像超分辨[J].光电工程,2013,40(11):76-82.(Ding Lan,Fu Zhizhong,Li Xiaofeng,et al. Image super resolution for multi-view mixture resolution based on non-uniform sampling restoration[J]. Opto-Electronic Engineering,2013,40(11):76-82.)
[10]Tang Yi,Yuan Yuan,Yan Pingkun,et al. Greedy regression in sparse coding space for single-image super-resolution[J]. Journal of Visual Communication&Image Representation,2013,24(2):148-159.
[11] Xu Hongteng,Zhai Guangtao,Yang Xiaokang. Single image superresolution with detail enhancement based on local fractal analysis of gradient[J]. IEEE Trans on Circuits&Systems for Video Technology,2013,23(10):1740-1754.
[12]Jiang Junjun,Hu Ruimin,Han Zhen,et al. Efficient single image super-resolution via graph-constrained least squares regression[J]. Multimedia Tools&Applications,2014,72(3):2573-2596.
[13]Zhu Zhiliang,Guo Fangda,Yu Hai,et al. Fast single image super-resolution via self-example learning and sparse representation[J]. IEEE Trans on Multimedia,2014,16(8):2178-2190.
[14]Chen Xiaoxuan,Qi Chun. Low-rank neighbor embedding for single image super-resolution[J]. Signal Processing,2014,94(1):6-22.
[15]Bevilacqua M,Roumy A,Guillemot C,et al. Single-image super-resolution via linear mapping of interpolated self-examples[J]. IEEE Trans on Image Processing,2014,23(12):5334-5437.
[16]Li Jinming,Gong Weiguo,Li Weihong,et al. Single-image super-resolution reconstruction based on global non-zero gradient penalty and non-local Laplacian sparse coding[J]. Digital Signal Processing,2014,26(1):101-112.
[17] Timofte R,De V,Van Gool L. Anchored neighborhood regression for fast example-based super-resolution[C]//Proc of IEEE International Conference on Computer Vision. Piscataway,NJ:IEEE Press,2013:1920-1927.
[18]Schulter S,Leistner C,Bischof H. Fast and accurate image upscaling with super-resolution forests[C]//Proc of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway,NJ:IEEE Press,2015:3791-3799.
[19]Peleg T,Elad M. A statistical prediction model based on sparse representations for single image super-resolution[J]. IEEE Trans on Image Processing,2014,23(6):2569-2582.
[20]Dong Chao,Loy C C,He Kaiming,et al. Learning a deep convolutional network for image super-resolution[C]//Proc of European Conference on Computer Vision. Berlin:Springer,2014:184-199.
[2] Yang M C,Wang Y C F. A self-learning approach to single image superresolution[J]. IEEE Trans on Multimedia,2013,15(3):498-508.
[3] Wang Lingfeng,Xiang Shiming,Meng Gaofeng,et al. Edge-directed single-image super-resolution via adaptive gradient magnitude self-interpolation[J]. IEEE Trans on Circuits&Systems for Video Technology,2013,23(8):1289-1299.
[4] Zhang Kaibing,Gao Xinbo,Tao Dacheng,et al. Single image superresolution with multiscale similarity learning[J]. IEEE Trans on Neural Networks&Learning Systems,2013,24(10):1648-1659.
[5] Rueda A,Malpica N,Romero E. Single-image super-resolution of brain MR images using overcomplete dictionaries[J]. Medical Image Analysis,2013,17(1):113-132.
[6] Zhang Kaibing,Tao Dacheng,Gao Xinbo,et al. Learning multiple linear mappings for efficient single image super-resolution[J]. IEEE Trans on Image Processing,2015,24(3):846-860.
[7]李银辉,吕晓琪,于荷峰.基于l1和l2混合范式的序列图像超分辨率重建[J].计算机应用,2015,35(3):840-843.(Li Yinhui,Lyu Xiaoqi,Yu Hefeng. Sequence images super-resolution reconstruction based on l1and l2mixed norm[J]. Journal of Computer Applications,2015,35(3):840-843.)
[8]郭琳,叶波,陈庆虎,等.融合几何变换相似块的序列图像超分辨率重建[J].光电工程,2015,42(8):79-85.(Guo Lin,Ye Bo,Chen Qinghu,et al. Super-resolution reconstruction of image sequences via fusing similar patches with geometric transformation[J].Opto-Electronic Engineering,2015,42(8):79-85.)
[9]丁兰,傅志中,李晓峰,等.使用非均匀采样复原理论进行裂纹修复的混合分辨率多视角图像超分辨[J].光电工程,2013,40(11):76-82.(Ding Lan,Fu Zhizhong,Li Xiaofeng,et al. Image super resolution for multi-view mixture resolution based on non-uniform sampling restoration[J]. Opto-Electronic Engineering,2013,40(11):76-82.)
[10]Tang Yi,Yuan Yuan,Yan Pingkun,et al. Greedy regression in sparse coding space for single-image super-resolution[J]. Journal of Visual Communication&Image Representation,2013,24(2):148-159.
[11] Xu Hongteng,Zhai Guangtao,Yang Xiaokang. Single image superresolution with detail enhancement based on local fractal analysis of gradient[J]. IEEE Trans on Circuits&Systems for Video Technology,2013,23(10):1740-1754.
[12]Jiang Junjun,Hu Ruimin,Han Zhen,et al. Efficient single image super-resolution via graph-constrained least squares regression[J]. Multimedia Tools&Applications,2014,72(3):2573-2596.
[13]Zhu Zhiliang,Guo Fangda,Yu Hai,et al. Fast single image super-resolution via self-example learning and sparse representation[J]. IEEE Trans on Multimedia,2014,16(8):2178-2190.
[14]Chen Xiaoxuan,Qi Chun. Low-rank neighbor embedding for single image super-resolution[J]. Signal Processing,2014,94(1):6-22.
[15]Bevilacqua M,Roumy A,Guillemot C,et al. Single-image super-resolution via linear mapping of interpolated self-examples[J]. IEEE Trans on Image Processing,2014,23(12):5334-5437.
[16]Li Jinming,Gong Weiguo,Li Weihong,et al. Single-image super-resolution reconstruction based on global non-zero gradient penalty and non-local Laplacian sparse coding[J]. Digital Signal Processing,2014,26(1):101-112.
[17] Timofte R,De V,Van Gool L. Anchored neighborhood regression for fast example-based super-resolution[C]//Proc of IEEE International Conference on Computer Vision. Piscataway,NJ:IEEE Press,2013:1920-1927.
[18]Schulter S,Leistner C,Bischof H. Fast and accurate image upscaling with super-resolution forests[C]//Proc of IEEE Conference on Computer Vision and Pattern Recognition. Piscataway,NJ:IEEE Press,2015:3791-3799.
[19]Peleg T,Elad M. A statistical prediction model based on sparse representations for single image super-resolution[J]. IEEE Trans on Image Processing,2014,23(6):2569-2582.
[20]Dong Chao,Loy C C,He Kaiming,et al. Learning a deep convolutional network for image super-resolution[C]//Proc of European Conference on Computer Vision. Berlin:Springer,2014:184-199.