基于卷积稀疏自编码的图像超分辨率重建

张秀; 周巍; 段哲民; 魏恒璐

doi:doi:10.3788/irla201948.0126005

红外与激光工程, 2019, 48 (1): 0126005, 网络出版: 2019-04-02

基于卷积稀疏自编码的图像超分辨率重建

Convolutional sparse auto-encoder for image super-resolution reconstruction

论文大纲

张秀周巍段哲民魏恒璐

作者单位

西北工业大学电子信息学院, 陕西西安 710129

稀疏编码卷积神经网络自编码器深度学习超分辨率重建 sparse coding convolutional neural network auto-encoder deep learning super-resolution reconstruction

摘要

针对卷积稀疏编码算法中特征映射的准确性的问题, 为了进一步提高图像超分辨率重建的的质量, 文中提出一种基于卷积稀疏自编码的图像超分辨率重建算法。该算法首先在预训练阶段利用稀疏自编码器对输入高低分辨率图像分别进行训练, 得到对应的图像稀疏特征表示; 然后再由卷积神经网络根据得到的稀疏系数共同训练相应的滤波器及特征映射函数并更新到最优解; 最后由高分辨率滤波器和对应的稀疏表示系数卷积求和, 得到高分辨率重建图像估计。实验结果显示, 改进算法的峰值信噪比(PSNR)结果较卷积稀疏编码算法提高了近0.1 dB, 有效提高了重建图像的质量。

Abstract

For the accuracy of feature maps in convolutional sparse coding algorithm, in order to further improve the quality of image super-resolution reconstruction, an image super-resolution(SR) reconstruction algorithm based on convolutional sparse auto-encoder was proposed in this paper. In this algorithm, firstly, the input images were pre-trained with sparse auto-encoder for obtaining the feature of LR and HR image; after that, the convolutional neural network trained the corresponding filters and feature mapping function and updated to the optimal solution according to the obtained sparse coefficients; finally, the summation of the convolutions of high-resolution(HR) filters and the corresponding feature maps could reconstruct the HR image. The experimental results show that the peak signal-to-noise ratio(PSNR) of the proposed algorithm is nearly 0.1 dB higher than the CSC algorithm, which improves the quality of reconstructed images.

参考文献

[1] Irani M, Peleg S. Improving resolution by image registration[J]. Cvgip Graphical Models & Image Processing, 1991, 53(3): 231-239.

[2] Lu X, Li X. Multiresolution imaging[J]. IEEE Trans Cybern, 2014, 44(1): 149-160.

[3] Hou H S, Andrews H C. Cubic spline for image interpolation and digital filtering[J]. IEEE Transaction on Signal Pressing, 1978, 26(6): 508-517.

[4] Tai Y W, Liu S, Brown M S, et al. Super resolution using edge prior and single image detail synthesis[C]//Computer Vision and Pattern Recognition. IEEE, 2010: 2400-2407.

[5] Sun J, Sun J, Xu Z, et al. Image super-resolution using gradient profile prior[C]//IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2008: 1-8.

[6] You X, Li Q, Tao D, et al. Local metric learning for exemplar-based object detection[J]. IEEE Transactions on Circuits & Systems for Video Technology, 2014, 24(8):1265-1276.

[7] Yu J, Gao X, Tao D, et al. A unified learning framework for single image super-resolution[J]. IEEE Trans Neural Netw Learn Syst, 2014, 25(4): 780-792.

[8] Ren J, Liu J, Guo Z. Context-aware sparse decomposition for image denoising and super-resolution[J]. IEEE Transactions on Image Processing, 2013, 22(4): 1454-1467.

[9] Romano Y, Protter M, Elad M. Single image interpolation via adaptive nonlocal sparsity-based modeling[J]. IEEE Transactions on Image Processing, 2014, 23(7): 3085.

[10] Yang J, Wright J, Huang T, et al. Image super-resolution as sparse representation of raw image patches[C]//Computer Vision and Pattern Recognition, 2008, IEEE Conference on. IEEE, 2008: 1-8.

[11] Wang S, Zhang D, Liang Y, et al. Semi-coupled dictionary learning with applications to image super-resolution and photo-sketch synthesis[C]//Computer Vision and Pattern Recognition. IEEE, 2012: 2216-2223.

[12] Zeiler M D, Krishnan D, Taylor G W, et al. Deconvolutional networks[C]//Computer Vision and Pattern Recognition. IEEE, 2010: 2528-2535.

[13] Gu S, Zuo W, Xie Q, et al. Convolutional sparse coding for image super-resolution[C]//IEEE International Conference on Computer Vision. IEEE, 2015: 1823-1831.

[14] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. American Association for the Advancement of Science, 2006, 313(5786): 504-507.

[15] Hinton G E, S.Osindero Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554.

[16] Lee H, Ekanadham C, Ng A Y. Sparse deep belief net model for visual area V2[C]//International Conference on Neural Information Processing Systems. Curran Associates Inc, 2007: 873-880.

[17] Sch 觟lkopf B, Platt J, Hofmann T. Greedy layer-wise training of deep networks[C]//Advances in Neural Information Processing Systems 19, Proceedings of the Twentieth Conference on Neural Information Processing Systems, 2007: 153-160.

[18] Lee Q V, Ngiam J, Coates A, et al. On optimization meyhods for deep learning[C]//Proceedings of the 28th International Methods for Deep Learning, 2011: 265-272.

[19] Kivinen J, Warmuth M K. Exponentiated gradient versus gradient descent for linear predictors[J]. Information and Computation, 1997, 132(1): 1-63.

[20] Wohlberg B. Efficient convolutional sparse coding[C]//IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2014: 7173-7177.

[21] Zhong L W, Kwok J T. Fast stochastic alternating direction method of multipliers[C]//International Conference on International Conference on Machine Learning, 2014: I-46.

[22] Kwang In Kim, Younghee Kwon. Single-image super-resolution using sparse regression and natural image prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(6): 1127.

张秀, 周巍, 段哲民, 魏恒璐. 基于卷积稀疏自编码的图像超分辨率重建[J]. 红外与激光工程, 2019, 48(1): 0126005. Zhang Xiu, Zhou Wei, Duan Zhemin, Wei Henglu. Convolutional sparse auto-encoder for image super-resolution reconstruction[J]. Infrared and Laser Engineering, 2019, 48(1): 0126005.

基于卷积稀疏自编码的图像超分辨率重建

关于本站 Cookie 的使用提示

全站搜索

基于卷积稀疏自编码的图像超分辨率重建

相关论文

相关资讯

关于本站 Cookie 的使用提示

全站搜索