整体车牌图像超分辨率重建研究
|
摘要
为了增强低分辨率车牌(LP)图像的整体重建效果,将深度学习思想应用到车牌图像超分辨率重建任务中,提出了一种基于卷积神经网络(CNN)的单幅整体车牌图像超分辨率重建(SR)算法(LPSRCNN)。对初始高分辨率车牌图像进行预处理,利用双三次插值法Bicubic对原始车牌图像进行下采样后再上采样得到和初始图像尺寸大小一样的低分辨率车牌图像作为网络的输入图像,通过训练好的卷积神经网络直接学习低分辨率车牌图像和高分辨率车牌图像之间的映射关系,利用该映射关系输出重建后的高分辨率整体车牌图像。将得到的研究结果与双三次插值法和字典学习方法进行比较,结果表明计算得到的峰值信噪比(PSNR)都高于这两种方法。将该方法应用到整体车牌图像超分辨率重建问题中,可以获取更丰富的细节信息,得到更好的视觉效果,达到了整体车牌图像超分辨率的增强任务。
In order to enhance the overall reconstruction effect of low-resolution license plate(LP) images,applying the idea of deep learning into the super-resolution reconstruction of license plate images,we propose a super-resolution algorithm based on convolutional neural network(CNN) for single license plate image,referred to as LPSRCNN. The original high-resolution license plate image is preprocessed,and the low-resolution license plate image with the same size as the original image is obtained after the Bicubic interpolation Bicubic sampling of the original license plate image. The trained convolutional neural network is used to directly learn the mapping relationship between low-resolution license plate image and high-resolution license plate image,and the reconstructed high-resolution overall license plate image is output based on the mapping relationship. Compared with the Bicubic interpolation method and the dictionary learning method,the results show that the PSNR is higher than that of the two methods. The method of deep learning is applied to the problem of super-resolution reconstruction of overall license plate image,which can obtain more details and better visual effect,achieving the task of enhancing the super resolution of the overall license plate image.
In order to enhance the overall reconstruction effect of low-resolution license plate(LP) images,applying the idea of deep learning into the super-resolution reconstruction of license plate images,we propose a super-resolution algorithm based on convolutional neural network(CNN) for single license plate image,referred to as LPSRCNN. The original high-resolution license plate image is preprocessed,and the low-resolution license plate image with the same size as the original image is obtained after the Bicubic interpolation Bicubic sampling of the original license plate image. The trained convolutional neural network is used to directly learn the mapping relationship between low-resolution license plate image and high-resolution license plate image,and the reconstructed high-resolution overall license plate image is output based on the mapping relationship. Compared with the Bicubic interpolation method and the dictionary learning method,the results show that the PSNR is higher than that of the two methods. The method of deep learning is applied to the problem of super-resolution reconstruction of overall license plate image,which can obtain more details and better visual effect,achieving the task of enhancing the super resolution of the overall license plate image.
引文
[1] 苏衡,周杰,张志浩.超分辨率图像重建方法综述[J].自动化学报,2013,39(8):1202-1213.
[2] 王春霞,苏红旗,范郭亮.图像超分辨率重建技术综述[J].计算机技术与发展,2011,21(5):124-127.
[3] 李志强,李永斌.车牌识别技术的发展及研究现状[J].科技信息,2012(5):110.
[4] 李波,曾致远,周建中,等.车牌识别系统研究与实现[J].计算机技术与发展,2006,16(6):10-11.
[5] 汤玉垚.复杂背景下车牌识别算法的研究[D].合肥:中国科学技术大学,2016.
[6] 顾弘,赵光宙,齐东莲,等.车牌识别中先验知识的嵌入及字符分割方法[J].中国图象图形学报,2010,15(5):749-756.
[7] LIANG Yudong,WANG Jinjun,ZHOU Sanping,et al.Incorporating image priors with deep convolutional neural networks for image super-resolution[J].Neurocomputing,2016,194:340-347.
[8] 李素梅,雷国庆,范如.基于卷积神经网络的深度图超分辨率重建[J].光学学报,2017,37(12):1210002.
[9] 黄凤,王晓明.增强的单幅图像自学习超分辨方法[J].计算机应用,2017,37(9):2636-2642.
[10] DONG Chao,LOY C C,HE Kaiming,et al.Learning a deep convolutional network for image super-resolution[C]//European conference on computer vision.Berlin:Springer,2014:184-199.
[11] OUYANG Wanli,WANG Xiaogang.Joint deep learning for pedestrian detection[C]//IEEE international conference on computer vision.Sydney,NSW,Australia:IEEE,2014:2056-2063.
[12] OUYANG Wanli,ZENG Xingyu,WANG Xiaogang,et al.DeepID-net:deformable deep convolutional neural networks for object detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017,39(7):1320-1334.
[13] ZHANG Ning,DONAHUE J,GIRSHICK R,et al.Part-based R-CNNs for fine-grained category detection[M]//13th European conference on computer vision.[s.l.]:Springer International Publishing,2014:834-849.
[14] SUN Yi,CHEN Yuheng,WANG Xiaogang,et al.Deep learning face representation by joint identification-verification[C]//Proceedings of the 27th international conference on neural information processing systems.Montreal,Canada:MIT Press,2014:1988-1996.
[15] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th international conference on neural information processing systems.Lake Tahoe,Nevada:Curran Associates Inc.,2012:1097-1105.
[16] NAIR V,HINTON G E.Rectified linear units improve restricted Boltzmann machines[C]//International conference on machine learning.Omnipress:[s.n.],2010:807-814.
[17] HOCHREITER S,BENGIO Y,FRASCONI P,et al.Gradient flow in recurrent nets:the difficulty of learning long-term dependencies[M].[s.l.]:IEEE Press,2001.
[18] 刘志州.基于字典学习的超分辨率图像重构[D].西安:西安电子科技大学,2011.
[19] 史国川,龚连友.基于系数复用和字典训练的图像超分辨率算法[J].计算机技术与发展,2018,28(3):114-117.
[2] 王春霞,苏红旗,范郭亮.图像超分辨率重建技术综述[J].计算机技术与发展,2011,21(5):124-127.
[3] 李志强,李永斌.车牌识别技术的发展及研究现状[J].科技信息,2012(5):110.
[4] 李波,曾致远,周建中,等.车牌识别系统研究与实现[J].计算机技术与发展,2006,16(6):10-11.
[5] 汤玉垚.复杂背景下车牌识别算法的研究[D].合肥:中国科学技术大学,2016.
[6] 顾弘,赵光宙,齐东莲,等.车牌识别中先验知识的嵌入及字符分割方法[J].中国图象图形学报,2010,15(5):749-756.
[7] LIANG Yudong,WANG Jinjun,ZHOU Sanping,et al.Incorporating image priors with deep convolutional neural networks for image super-resolution[J].Neurocomputing,2016,194:340-347.
[8] 李素梅,雷国庆,范如.基于卷积神经网络的深度图超分辨率重建[J].光学学报,2017,37(12):1210002.
[9] 黄凤,王晓明.增强的单幅图像自学习超分辨方法[J].计算机应用,2017,37(9):2636-2642.
[10] DONG Chao,LOY C C,HE Kaiming,et al.Learning a deep convolutional network for image super-resolution[C]//European conference on computer vision.Berlin:Springer,2014:184-199.
[11] OUYANG Wanli,WANG Xiaogang.Joint deep learning for pedestrian detection[C]//IEEE international conference on computer vision.Sydney,NSW,Australia:IEEE,2014:2056-2063.
[12] OUYANG Wanli,ZENG Xingyu,WANG Xiaogang,et al.DeepID-net:deformable deep convolutional neural networks for object detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017,39(7):1320-1334.
[13] ZHANG Ning,DONAHUE J,GIRSHICK R,et al.Part-based R-CNNs for fine-grained category detection[M]//13th European conference on computer vision.[s.l.]:Springer International Publishing,2014:834-849.
[14] SUN Yi,CHEN Yuheng,WANG Xiaogang,et al.Deep learning face representation by joint identification-verification[C]//Proceedings of the 27th international conference on neural information processing systems.Montreal,Canada:MIT Press,2014:1988-1996.
[15] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th international conference on neural information processing systems.Lake Tahoe,Nevada:Curran Associates Inc.,2012:1097-1105.
[16] NAIR V,HINTON G E.Rectified linear units improve restricted Boltzmann machines[C]//International conference on machine learning.Omnipress:[s.n.],2010:807-814.
[17] HOCHREITER S,BENGIO Y,FRASCONI P,et al.Gradient flow in recurrent nets:the difficulty of learning long-term dependencies[M].[s.l.]:IEEE Press,2001.
[18] 刘志州.基于字典学习的超分辨率图像重构[D].西安:西安电子科技大学,2011.
[19] 史国川,龚连友.基于系数复用和字典训练的图像超分辨率算法[J].计算机技术与发展,2018,28(3):114-117.