基于贝塔过程联合字典学习的图像超分辨重建
|
摘要
为了提高图像超分辨效果,针对以往稀疏字典超分辨算法仅适用于单特征空间的问题,提出基于贝塔过程联合字典学习(BPJDL)的图像超分辨重建(SRR)方法。首先,根据图像退化模型生成训练样本图像,分别对高、低分辨率图像进行7×7分块,并利用吉布斯采样对图像块进行采样,生成字典训练样本。然后,依据贝塔过程先验模型,建立连接高、低分辨率图像空间的双参数联合稀疏字典,将字典稀疏系数分解为系数权值和字典原子的乘积,通过训练和更新字典,得到同时适用于两个特征空间的字典映射矩阵。最后,进行图像超分辨稀疏重构。实验结果表明:本文方法能以更小尺寸的稀疏字典重建超分辨图像,与当前最先进的稀疏表示超分辨算法相比,结果图像主观视觉上纹理细节信息更丰富,客观评价参数峰值信噪比(PSNR)提高约1.5 dB,结构相似性(SSIM)提高约0.02,超分辨重建时间降低约50 s。
Aiming to solve the problem that the learning dictionary can only apply to single feature space, an image super-resolution reconstruction(SRR) method is proposed to improve the image SRR effect, which is based on the Beta process joint dictionary learning(BPJDL). Firstly, training sample images are generated according to image degradation model, and high-low resolution images are respectively divided into image patches size of 7×7. Gibbs sampling is used to generate the dictionary training samples. Then, according to the prior model of BP, a double-parameter joint training dictionary is established to connect the high-low resolution image space. The dictionary sparse coefficients are expressed as a multiplication of coefficient weights and dictionary atoms. After training and updating,a learning dictionary is obtained applying to coupled feature spaces. Finally, image SRR is performed. Experiment results show that the method proposed here can reconstruct the higher quality image with smaller sparse dictionary. At the same time, compared with the most advanced sparse representation SRR algorithm, the result images contain more texture details subjectively, and its objective parameters are better. Its PSNR is increased by about 1.5 dB, SSIM is increased by about 0.02, SRR time is reduced by 50 s.
Aiming to solve the problem that the learning dictionary can only apply to single feature space, an image super-resolution reconstruction(SRR) method is proposed to improve the image SRR effect, which is based on the Beta process joint dictionary learning(BPJDL). Firstly, training sample images are generated according to image degradation model, and high-low resolution images are respectively divided into image patches size of 7×7. Gibbs sampling is used to generate the dictionary training samples. Then, according to the prior model of BP, a double-parameter joint training dictionary is established to connect the high-low resolution image space. The dictionary sparse coefficients are expressed as a multiplication of coefficient weights and dictionary atoms. After training and updating,a learning dictionary is obtained applying to coupled feature spaces. Finally, image SRR is performed. Experiment results show that the method proposed here can reconstruct the higher quality image with smaller sparse dictionary. At the same time, compared with the most advanced sparse representation SRR algorithm, the result images contain more texture details subjectively, and its objective parameters are better. Its PSNR is increased by about 1.5 dB, SSIM is increased by about 0.02, SRR time is reduced by 50 s.
引文
[ 1] Glasner D,Bagon S,Irani M.Super-resolution from a single image[C].In:IEEE International Conference on Computer Vision,Kyoto,Japan,2009.349-356
[ 2] 朱福珍,王晓飞,丁群,等.三级训练BP神经网络遥感图像超分辨重建[J].光学精密工程,2015,23(10):208-215
[ 3] 吴强,王国林,许荣庆.两种超分辨ISAR成像算法[J].高技术通讯,1999,9(10):25-29
[ 4] Freeman W T,Jones T R,Pasztor E C.Exampled-based super-resolution[J].IEEE Computer Graphics and Applications,2002,22(5):56-65
[ 5] Ling F,Zhang Y,Foody G M,et al.Learning-based super-resolution land cover mapping[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(7):3794-3810
[ 6] Yuan Y,Zheng X,Lu X.Hyperspectral image super-resolution by transfer learning[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2017,10(5):1963-1974
[ 7] Chang H,Yeung D Y,Xiong Y.Super-resolution through neighbor embedding[C].In:Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Washington,USA,2004.275-282
[ 8] Sun J,Zheng N N,Tao H,et al.Image hallucination with primal sketch priors[C].In:IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Madison,USA,2003.29-36
[ 9] Xie Q L,Chen H,Cao H M.Improved example-based single-based single-image super-resolution[C].In:The 3rd International Congress on Image and Signal Processing,Yantai,China,2010.1204-1207
[10] Zhao L,Sun Q,Zhang Z.Single image super-resolution based on deep learning features and dictionary model[J].IEEE Access,2017,5(99):17126-17135
[11] Turkan M,Guillemot C.Online dictionaries for image prediction[C].In:IEEE International Conference on Image Processing,Brussels,Belgium,2011.293-296
[12] Scholhopf B,Platt J,Hofmann T.Efficient sparse coding algorithms[C].In:Conference on Neural Information Processing Systems,Vancouver,Canada,2006.801-808
[13] Zeyde R,Michael E,Matan P.On single image scale-up using sparse-representations [C].In:International Conference on Curves and Surfaces,Paris,France,2010.711-730
[14] Yang J,Wright J,Huang T S,et al.Image super-resolution via sparse representation [J].IEEE Transactions on Image Processing,2010,19(11):2861-2873
[15] Yang J C,Wang Z W,Lin Z.Bilevel sparse coding for coupled feature spaces[C].In:IEEE Conference on Computer Vision and Pattern Recognition,Providence,USA,2012.2360-2367
[16] Yang J,Wright J,Huang T,et al.Image super-resolution as sparse representation of raw image patches[C].In:IEEE Conference on Computer Vision and Pattern Recognition,Anchorage,USA,2008.1-8
[17] He L,Qi H,Zaretaki R.Beta process joint dictionary learning for coupled feature spaces with application to single image super-resolution[C].In:IEEE Computer Vision and Pattern Recognition,Portland,USA,2013.345-352
[18] Zhou M,Chen H,Paisley J,et al.Nonparametric Bayesian dictionary learning for analysis of noisy and incomplete images[J].IEEE Transactions on Image Processing,2012,21(1):130-144
[19] 李绍滨,赵淑清,辛海华.利用自适应阵对空间信号测向[J].高技术通讯,2001,11(8):69-72
[20] Song P F,Joshua D T,Armando M,et al.Improved super-resolution ultrasound microvessel imaging with spatiotemporal nonlocal means filtering and bipartite graph-based microbubble tracking[J].IEEE Transactions on Ultrasonics,Ferroelectrics and Frequency Control,2018,65(2):149-167
[ 2] 朱福珍,王晓飞,丁群,等.三级训练BP神经网络遥感图像超分辨重建[J].光学精密工程,2015,23(10):208-215
[ 3] 吴强,王国林,许荣庆.两种超分辨ISAR成像算法[J].高技术通讯,1999,9(10):25-29
[ 4] Freeman W T,Jones T R,Pasztor E C.Exampled-based super-resolution[J].IEEE Computer Graphics and Applications,2002,22(5):56-65
[ 5] Ling F,Zhang Y,Foody G M,et al.Learning-based super-resolution land cover mapping[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(7):3794-3810
[ 6] Yuan Y,Zheng X,Lu X.Hyperspectral image super-resolution by transfer learning[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2017,10(5):1963-1974
[ 7] Chang H,Yeung D Y,Xiong Y.Super-resolution through neighbor embedding[C].In:Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Washington,USA,2004.275-282
[ 8] Sun J,Zheng N N,Tao H,et al.Image hallucination with primal sketch priors[C].In:IEEE Computer Society Conference on Computer Vision and Pattern Recognition,Madison,USA,2003.29-36
[ 9] Xie Q L,Chen H,Cao H M.Improved example-based single-based single-image super-resolution[C].In:The 3rd International Congress on Image and Signal Processing,Yantai,China,2010.1204-1207
[10] Zhao L,Sun Q,Zhang Z.Single image super-resolution based on deep learning features and dictionary model[J].IEEE Access,2017,5(99):17126-17135
[11] Turkan M,Guillemot C.Online dictionaries for image prediction[C].In:IEEE International Conference on Image Processing,Brussels,Belgium,2011.293-296
[12] Scholhopf B,Platt J,Hofmann T.Efficient sparse coding algorithms[C].In:Conference on Neural Information Processing Systems,Vancouver,Canada,2006.801-808
[13] Zeyde R,Michael E,Matan P.On single image scale-up using sparse-representations [C].In:International Conference on Curves and Surfaces,Paris,France,2010.711-730
[14] Yang J,Wright J,Huang T S,et al.Image super-resolution via sparse representation [J].IEEE Transactions on Image Processing,2010,19(11):2861-2873
[15] Yang J C,Wang Z W,Lin Z.Bilevel sparse coding for coupled feature spaces[C].In:IEEE Conference on Computer Vision and Pattern Recognition,Providence,USA,2012.2360-2367
[16] Yang J,Wright J,Huang T,et al.Image super-resolution as sparse representation of raw image patches[C].In:IEEE Conference on Computer Vision and Pattern Recognition,Anchorage,USA,2008.1-8
[17] He L,Qi H,Zaretaki R.Beta process joint dictionary learning for coupled feature spaces with application to single image super-resolution[C].In:IEEE Computer Vision and Pattern Recognition,Portland,USA,2013.345-352
[18] Zhou M,Chen H,Paisley J,et al.Nonparametric Bayesian dictionary learning for analysis of noisy and incomplete images[J].IEEE Transactions on Image Processing,2012,21(1):130-144
[19] 李绍滨,赵淑清,辛海华.利用自适应阵对空间信号测向[J].高技术通讯,2001,11(8):69-72
[20] Song P F,Joshua D T,Armando M,et al.Improved super-resolution ultrasound microvessel imaging with spatiotemporal nonlocal means filtering and bipartite graph-based microbubble tracking[J].IEEE Transactions on Ultrasonics,Ferroelectrics and Frequency Control,2018,65(2):149-167