基于残差神经网络的高强度运动超分辨率图像重构
|
摘要
针对实际拍摄的亚像素信息较少的低分辨率运动图像,重构图像通常较为模糊,甚至不能分辨。为此,提出一种新的基于残差神经网络的高强度运动超分辨率图像重构方法。令沿运动方向的亮度保持恒定,通过光流场匹配实现高强度运动图像的运动估计;根据运动估计结果和超分辨率重构的基本思想,将BP神经网络看作残差神经网络的基础建立残差神经网络,对残差神经网络进行训练,参照训练样本将经插值法放大若干倍的待重构高强度运动图像作为输入,将高分辨率图像和输入图像间的残差作为输出,把输入和输出累加获取超分辨率图像,实现若干放大倍数高强度运动超分辨率图像的重构。实验结果表明,所提方法运动估计准确,重构图像清晰、质量佳。
The reconstructed image is usually blurred and even can not be resolved in view of the low resolution moving image with less sub-pixel information. To this end,a new method of high-intensity motion super-resolution image reconstruction based on residual neural network is proposed. The brightness along the direction of motion is kept constant by optical flow field matching high intensity motion image motion estimation; motion estimation results and according to the basic idea of super-resolution reconstruction based on BP neural network,the neural network is established as the residual error of neural network,the training of residual neural network training samples,reference by interpolation method amplified several times to reconstruct high intensity motion image as input,the residual high resolution image and the input image as the output,input and output to the cumulative acquisition of super resolution image,to achieve a number of high magnification image super resolution reconstruction of motion intensity. The experimental results show that the proposed method is accurate in motion estimation,and the reconstructed image is clear and good in quality.
The reconstructed image is usually blurred and even can not be resolved in view of the low resolution moving image with less sub-pixel information. To this end,a new method of high-intensity motion super-resolution image reconstruction based on residual neural network is proposed. The brightness along the direction of motion is kept constant by optical flow field matching high intensity motion image motion estimation; motion estimation results and according to the basic idea of super-resolution reconstruction based on BP neural network,the neural network is established as the residual error of neural network,the training of residual neural network training samples,reference by interpolation method amplified several times to reconstruct high intensity motion image as input,the residual high resolution image and the input image as the output,input and output to the cumulative acquisition of super resolution image,to achieve a number of high magnification image super resolution reconstruction of motion intensity. The experimental results show that the proposed method is accurate in motion estimation,and the reconstructed image is clear and good in quality.
引文
1杜伟男,胡永利,孙艳丰.基于残差字典学习的图像超分辨率重建方法.北京工业大学学报,2017;43(1):43-48Du Weinan,Hu Yongli,Sun Yanfeng.Image super-resolution reconstruction based on residual dictionary learning.Journal of Beijing University of Technology,2017;43(1):43-48
2陈烽.超分辨率数字图像特征提取及重构方法研究.科学技术与工程,2017;17(11):255-259Chen Feng.Super resolution digital image feature extraction and reconstruction methods.Science Technology and Engineering,2017;17(11):255-259
3 Bahy R M,Salama G I,Mahmoud T A.Adaptive regularizationbased super resolution reconstruction technique for multi-focus lowresolution images.Signal Processing,2014;103(1):155-167
4常志国,郭茹侠,李晶,等.基于稀疏表示和近邻嵌入的图像超分辨率重构.计算机测量与控制,2016;24(5):173-177Chang Zhiguo,Guo Ruxia,Li Jing,et al.Image super-resolution reconstruction based on sparse representation and neighbor embedding.Computer Measurement&Control,2016;24(5):173-177
5刘晨羽,蒋云飞,李学明.基于卷积神经网的单幅图像超分辨率重建算法.计算机辅助设计与图形学学报,2017;29(9):1643-1649Liu Chenyu,Jiang Yunfei,Li Xueming.Single image super-resolution reconstruction using convolutional neural networks.Journal of Computer-Aided Design&Computer Graphics,2017;29(9):1643-1649
6巫乾军,孙艳丰,赵璐.稀疏表示的深度图像超分辨率重建研究与仿真.计算机仿真,2017;34(5):234-237Wu Qianjun,Sun Yanfeng,Zhao Lu.Depth image super-resolution reconstruction of the sparse representation and simulation.Computer Simulation,2017;34(5):234-237
7 Dzyubachyk O,Tao Q,Poot D H,et al.Super-resolution reconstruction of late gadolinium-enhanced MRI for improved myocardial scar assessment.Journal of Magnetic Resonance Imaging,2015;42(1):160-167
8徐冉,张俊格,黄凯奇.利用双通道卷积神经网络的图像超分辨率算法.中国图象图形学报,2016;21(5):556-564Xu Ran,Zhang Junge,Huang Kaiqi.Image super-resolution using two-channel convolutional neural networks.Journal of Image and Graphics,2016;21(5):556-564
9包莹莹,王华君,徐燕华,等.基于稀疏编码和随机森林的多帧图像超分辨率算法.电子设计工程,2017;25(8):158-162Bao Yingying,Wang Huajun,Xu Yanhua,et al.A multi-frame image super resolution algorithm using sparse coding and random forest.Electronic Design Engineering,2017;25(8):158-162
10曾凯,丁世飞.图像超分辨率重建的研究进展.计算机工程与应用,2017;53(16):29-35Zeng Kai,Ding Shifei.Advances in image super-resolution reconstruction.Computer Engineering and Applications,2017;53(16):29-35
11郭丙华,岑志松.小波去噪和神经网络相融合的超分辨率图像重建.激光杂志,2016;37(2):61-64Guo Binghua,Cen Zhisong.Super resolution image reconstruction based on wavelet denoising and neural network.Laser Journal,2016;37(2):61-64
12禹建丽,黄鸿琦.基于BP神经网络的复杂过程参数优化方法研究.科技通报,2017;33(8):114-118Yu Jianli,Huang Hongqi.Study on parameter optimization method of complex process based on BP neural network.Bulletin of Science and Technology,2017;33(8):114-118
13 Li J,Gong W,Li W,et al.Single-image super-resolution reconstruction based on global non-zero gradient penalty and non-local Laplacian sparse coding.Digital Signal Processing,2014;26(1):101-112
14欧阳宁,曾梦萍,林乐平.基于并列卷积神经网络的超分辨率重建.计算机应用,2017;37(4):1174-1178Ouyang Ning,Zeng Mengping,Lin Leping.Parallel convolutional neural network for super-resolution reconstruction.Journal of Computer Applications,2017;37(4):1174-1178
15胡长胜,詹曙,吴从中.基于深度特征学习的图像超分辨率重建.自动化学报,2017;43(5):814-821Hu Changsheng,Zhan Shu,Wu Congzhong.Image super-resolution based on deep learning features.Acta Automatica Sinica,2017;43(5):814-821
16龙法宁,朱晓姝,胡春娇.基于深层卷积网络的单幅图像超分辨率重建模型.广西科学,2017;24(3):231-235Long Faning,Zhu Xiaoshu,Hu Chunjiao.Single image super-resolution restoration model using deep convolutional networks.Guangxi Sciences,2017;24(3):231-235
2陈烽.超分辨率数字图像特征提取及重构方法研究.科学技术与工程,2017;17(11):255-259Chen Feng.Super resolution digital image feature extraction and reconstruction methods.Science Technology and Engineering,2017;17(11):255-259
3 Bahy R M,Salama G I,Mahmoud T A.Adaptive regularizationbased super resolution reconstruction technique for multi-focus lowresolution images.Signal Processing,2014;103(1):155-167
4常志国,郭茹侠,李晶,等.基于稀疏表示和近邻嵌入的图像超分辨率重构.计算机测量与控制,2016;24(5):173-177Chang Zhiguo,Guo Ruxia,Li Jing,et al.Image super-resolution reconstruction based on sparse representation and neighbor embedding.Computer Measurement&Control,2016;24(5):173-177
5刘晨羽,蒋云飞,李学明.基于卷积神经网的单幅图像超分辨率重建算法.计算机辅助设计与图形学学报,2017;29(9):1643-1649Liu Chenyu,Jiang Yunfei,Li Xueming.Single image super-resolution reconstruction using convolutional neural networks.Journal of Computer-Aided Design&Computer Graphics,2017;29(9):1643-1649
6巫乾军,孙艳丰,赵璐.稀疏表示的深度图像超分辨率重建研究与仿真.计算机仿真,2017;34(5):234-237Wu Qianjun,Sun Yanfeng,Zhao Lu.Depth image super-resolution reconstruction of the sparse representation and simulation.Computer Simulation,2017;34(5):234-237
7 Dzyubachyk O,Tao Q,Poot D H,et al.Super-resolution reconstruction of late gadolinium-enhanced MRI for improved myocardial scar assessment.Journal of Magnetic Resonance Imaging,2015;42(1):160-167
8徐冉,张俊格,黄凯奇.利用双通道卷积神经网络的图像超分辨率算法.中国图象图形学报,2016;21(5):556-564Xu Ran,Zhang Junge,Huang Kaiqi.Image super-resolution using two-channel convolutional neural networks.Journal of Image and Graphics,2016;21(5):556-564
9包莹莹,王华君,徐燕华,等.基于稀疏编码和随机森林的多帧图像超分辨率算法.电子设计工程,2017;25(8):158-162Bao Yingying,Wang Huajun,Xu Yanhua,et al.A multi-frame image super resolution algorithm using sparse coding and random forest.Electronic Design Engineering,2017;25(8):158-162
10曾凯,丁世飞.图像超分辨率重建的研究进展.计算机工程与应用,2017;53(16):29-35Zeng Kai,Ding Shifei.Advances in image super-resolution reconstruction.Computer Engineering and Applications,2017;53(16):29-35
11郭丙华,岑志松.小波去噪和神经网络相融合的超分辨率图像重建.激光杂志,2016;37(2):61-64Guo Binghua,Cen Zhisong.Super resolution image reconstruction based on wavelet denoising and neural network.Laser Journal,2016;37(2):61-64
12禹建丽,黄鸿琦.基于BP神经网络的复杂过程参数优化方法研究.科技通报,2017;33(8):114-118Yu Jianli,Huang Hongqi.Study on parameter optimization method of complex process based on BP neural network.Bulletin of Science and Technology,2017;33(8):114-118
13 Li J,Gong W,Li W,et al.Single-image super-resolution reconstruction based on global non-zero gradient penalty and non-local Laplacian sparse coding.Digital Signal Processing,2014;26(1):101-112
14欧阳宁,曾梦萍,林乐平.基于并列卷积神经网络的超分辨率重建.计算机应用,2017;37(4):1174-1178Ouyang Ning,Zeng Mengping,Lin Leping.Parallel convolutional neural network for super-resolution reconstruction.Journal of Computer Applications,2017;37(4):1174-1178
15胡长胜,詹曙,吴从中.基于深度特征学习的图像超分辨率重建.自动化学报,2017;43(5):814-821Hu Changsheng,Zhan Shu,Wu Congzhong.Image super-resolution based on deep learning features.Acta Automatica Sinica,2017;43(5):814-821
16龙法宁,朱晓姝,胡春娇.基于深层卷积网络的单幅图像超分辨率重建模型.广西科学,2017;24(3):231-235Long Faning,Zhu Xiaoshu,Hu Chunjiao.Single image super-resolution restoration model using deep convolutional networks.Guangxi Sciences,2017;24(3):231-235