基于残差字典学习的图像超分辨率重建方法

杜伟男; 胡永利; 孙艳丰

doi:10.11936/bjutxb2016060049

基于残差字典学习的图像超分辨率重建方法

doi: 10.11936/bjutxb2016060049

北京工业大学多媒体与智能软件技术北京市重点实验室, 北京 100124

基金项目: 国家自然科学基金资助项目(61370119)

详细信息

作者简介:
作者简介: 杜伟男(1989—), 女, 博士研究生, 主要从事图像处理、模式识别方面的研究, E-mail:duweinan@emails.bjut.edu.cn

中图分类号: TP391.41
计量
- 文章访问数: 101
- HTML全文浏览量: 66
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2016-06-16
- 网络出版日期: 2022-09-09
- 刊出日期: 2017-01-01

Image Super-resolution Reconstruction Based on Residual Dictionary Learning

Beijing Key Laboratory of Multimedia and Intelligent Software Technology, Beijing University of Technology, Beijing 100124, China

摘要

摘要: 为了提高图像重建质量,在保留图像空间结构信息的同时恢复更多图像高频信息,提出一种基于二维可分离字典和残差字典的图像超分辨率重建方法. 不同于传统的基于一维字典的超分辨重建方法,二维字典直接利用图像的二维矩阵表示,因此,可以保持图像的空间结构信息,减少字典参数的数量,节省存储空间. 为了更好地恢复图像高频信息,在二维可分离字典重建图像基础上,引入残差字典,重建边缘等高频信息,两类字典各有侧重,二者结合可得到更高质量的超分辨率重建图像. 在典型的公共图像集上的实验证明了提出的结合二维可分离字典和残差字典的图像超分辨重建方法的有效性和优越性.
- 超分辨率重建 /
- 二维可分离字典 /
- 残差字典
Abstract: In this paper, an image super-resolution method was proposed based on 2D separable dictionary and residual dictionary to improve the quality of reconstructed image and to preserve space information as well as high-frequency information. Unlike the conventional super-resolution method based on 1D dictionary, the 2D separable dictionary was constructed from a set of training images in 2D matrix form without vectorization, so it had the capacity of learning the inherent texture structure of images. Additionally, it was a pair of compact dictionaries that were of smaller size and less storage space compared with the conventional dictionary. To restore more high-frequency information, the residual dictionary was introduced based on the reconstructed images with the 2D separable dictionary, which captured the high-frequency information of edges, angles and corners in images. Combining the two dictionary learning procedures into one framework, the proposed method was expected to synthesize high resolution images with high quality. The proposed algorithm was tested on public natural image set. The experiment results show that the proposed image super-resolution method based on 2D separable dictionary and residual dictionary is effective and superior.
- super-resolution /
- 2D separable dictionary /
- residual dictionary
The authors have declared that no competing interests exist.

HTML全文

图 1 整体框架图

Figure 1. Whole frame chat

下载: 全尺寸图片幻灯片

图 2 训练集图片示例

Figure 2. Samples of training images collected from Internet

下载: 全尺寸图片幻灯片

图 3 选择的测试图片

Figure 3. Some test images

下载: 全尺寸图片幻灯片

图 4 字典原子可视化显示

Figure 4. Visualizations of dictionary atoms

下载: 全尺寸图片幻灯片

图 5 图像重建结果对比(自左至右为双三次插值法^[3]、一维字典法^[11]、一维+残差法、二维字典法、二维+残差法、双重字典法^[19])

Figure 5. Comparison of reconstructed images(from left to right: bi-cubic method^[3], 1D-dictionary method^[11], 1D-dictionary-residual method, 2D-dictionary method, 2D-dictionary-residual method, Dual-dictionary method^[19])

下载: 全尺寸图片幻灯片

表 1 实验结果PSNR对比

Table 1. Experiment results of PSNR comparison

方法	字典参数个数	Lena	Pepper	Butterfly	Baboon
双三次插值法^[3]		34.13	33.06	27.06	24.95
一维字典法^[11]	131072	35.87	34.19	29.69	25.70
一维+残差法	196608	35.93	34.21	29.81	25.58
二维字典法	512	35.25	33.75	28.52	25.61
二维+残差法	66048	35.44	33.89	28.78	25.64
双重字典法^[19]	131072	34.72	33.71	28.67	24.07

下载: 导出CSV

参考文献(20)

[1]	PARK S, PARK M, KANG M.Super-resolution image reconstruction: a technical overview[J]. IEEE Signal Processing Magazine, 2003, 20(3): 21-36.
[2]	YANG J, HUANG T.Super-resolution imaging[M]. Boca Raton: CRC Press, 2010: 3-35.
[3]	ZHANG X, WU X.Image interpolation by adaptive 2-D autoregressive modeling and soft-decision estimation[J]. IEEE Transactions on Image Processing, 2008, 17(6): 887-896.
[4]	TAKEDA H, FARSIU S, MILANFAR P.Kernel regression for image processing and reconstruction[J]. IEEE Transactions on Image Processing, 2007, 15(2): 349-366.
[5]	FAUSIU S, ROBINSON M D, ELAD M, et al.Fast and robust multiframe super-resolution[J]. IEEE Transactions on Image Processing, 2004, 13(10): 1327-1344.
[6]	TIPPING M E, BISHOP C M.Bayesian image super-resolution[J]. Advances in Neural Information Processing Systems, 2002, 15: 1303-1310.
[7]	FATTAL R. Image upsampling via imposed edge statistics[J]. ACM Transactions on Graphics, 2007, 26(3): Article 95.
[8]	CHANG H, YEUNG D, XIONG Y.Super-resolution through neighbor embedding[C]//IEEE Conference on Computer Vision and Pattern Recognition. Washington, D C: IEEE, 2004: 275-282.
[9]	SUN J, XU Z, SHUM H.Image super-resolution using gradient profile prior[C]//IEEE Conference on Computer Vision and Pattern Recognition. Anchorage: IEEE, 2008: 1-8.
[10]	FERNANDEZ-GRANDA C, CANDES E J.Super-resolution via transform-invariant group-sparse regularizetion[C]//IEEE International Conference on Computer Vision. Sydney: IEEE, 2013: 3336-3343.
[11]	YANG J, WRIGHT J, HUANG T, et al.Image super-resolution via sparse representation[J]. IEEE Transactions on Image Processing, 2010, 19(11): 2861-2873.
[12]	GAO X, ZHANG K, TAO D, et al.Joint learning for single-image super-resolution via a coupled constraint[J]. IEEE Transactions on Image Processing, 2012, 21(2): 469-480.
[13]	YANG J, WANG Z, LIN Z, et al.Coupled dictionary training for image super resolution[J]. IEEE Transactions on Image Processing, 2012, 21(8): 3467-3478.
[14]	YANG M, YANG C.Fast direct superresolution by simple functions[C]//IEEE International Conference on Computer Vision. Sydney: IEEE, 2013: 561-568.
[15]	ZHANG K, TAO D, GAO X, et al.Learning multiple linear mappings for efficient single image super-resolution[J]. IEEE Transactions on Image Processing, 2015, 24(3): 846-861.
[16]	WANG L, WU H, PAN C.Fast image upsampling via the displacement field[J]. IEEE Transactions on Image Processing, 2014, 23(12): 5123-5135.
[17]	HAWE S, SEIBERT M, KLEINSTEUBER M.Separable dictionary learning[C]//IEEE International Conference on Computer Vision and Pattern Recognition. Portland: IEEE, 2013, 9(4): 438-445.
[18]	CAIAFA C, CICHOCKI A.Computing sparse representations of multidimensional signals using Kronecker bases[J]. Neural Computation, 2013, 25(1): 186-220.
[19]	ZHANGJZHAOCXIONGRet al.Image super-resolution via dual-dictionary learning and sparse representationIEEE International Symposium on Circuits and Systems. Seoul: IEEE201216881691 ZHANG J, ZHAO C, XIONG R, et al.Image super-resolution via dual-dictionary learning and sparse representation[C]//IEEE International Symposium on Circuits and Systems. Seoul: IEEE, 2012: 1688-1691.
[20]	AHARON M, ELAD M, BRUCKSTEIN A.K-SVD: an algorithm for designing overcomplete dictionaries for sparse representation[J]. IEEE Transactions on Signal Processing, 2006, 54(11): 4311-4322.