摘要
提出了一种基于生成器网络的遥感图像超分辨率方法。该方法对网络的权重随机初始化,无需使用数据集进行预训练。输入待修复的低分辨率图像,通过迭代训练网络,生成网络输出边缘更为清晰的超分辨率图像。对遥感图像上复杂纹理的修复也达到了更清晰的预期效果。实验结果表明,生成网络能在遥感图像超分辨率上达到经过大量数据集训练的卷积神经网络相似甚至更好的表现。
A remote sensing image super-resolution method based on generator network is proposed. This method does not require data sets for pre-training,and weights in it are randomly initialized. Low resolution image is imported into the generator network,which can output sharper high resolution image through iterative training. The effect of generator network on the restoration of complex texture on remote sensing images has achieved the expected results. Experimental results show that non-preconditioned generation network can achieve the similar or even better performance of the convolutional neural network trained by a large number of data sets at super-resolution of images.
引文
[1]樊学武,赵惠,易红伟.光学遥感器成像品质的主被动提升技术[J].航天返回与遥感,2013,34(3):16-25.
[2] Keys R G. Cubic convolution interpolation for digital image processing[J]. IEEE Transactions on Acoustics Speech&Signal Processing,2003,29(6):1153-1160.
[3]刘丁峰.超分辨率图像复原技术综述[J].软件导刊,2009,8(12):183-185.
[4]郭晓,谭文安.基于级联深度卷积神经网络的高性能图像超分辨率重构[J].计算机应用,2017,37(11):3124-3127,3144.
[5] Zhang Y,Du Y,Ling F,et al. Example-based super-resolution land cover mapping using support vector regression[J]. IEEE Journal of Selected Topics in Applied Earth Observations&Remote Sensing,2014,7(4):1271-1283.
[6]浦剑,张军平.基于词典学习和稀疏表示的超分辨率方法[J].模式识别与人工智能,2010,23(3):335-340.
[7] Dong C,Chen C L,He K,et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2016,38(2):295-307.
[8] Ledig C,Theis L,Huszár F,et al. Photo-realistic single image super-resolution using a generative adversarial network[C]//Computer Vision and Pattern Recognition. Honolulu, Hawaii:IEEE,2017:105-114.
[9]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(1):1-17.
[10] Zhang C,Bengio S,Hardt M,et al. Understanding deep learning requires rethinking generalization[DB/OL].(2017-2-26)[2018-9-20]. https://arxiv. org/pdf/1611. 03530v2. pdf.
[11] Ulyanov D,Vedaldi A,Lempitsky V. Deep Image Prior[DB/OL].(2018-4-5)[2018-9-20]. https://arxiv. org/pdf/1711.10925v3. pdf.
[12] A. M,A. V. Understanding deep image representations by inverting them[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). New York:IEEE Press,2015:5188-5196.
[13] J. K,J. K L,K. M L. Accurate Image Super-Resolution Using Very Deep Convolutional Networks[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Las Vegas:IEEE Press,2016:1646-1654.
[14]江静,张雪松.图像超分辨率重建算法综述[J].红外技术,2012(1):24-30.
[15] Kingma D P,Ba J. Adam:A method for stochastic optimization[DB/OL].(2017-1-30)[2018-9-20]. https://arxiv. org/pdf/1412. 6980. pdf.
[16] Bahdanau D,Cho K,Bengio Y. Neural machine translation by jointly learning to align and translate[DB/OL].(2016-5-19)[2018-9-20]. https://arxiv. org/pdf/1409. 0473. pdf.
[17]陆敏俊,王慈.基于相关性的JPEG高压缩图像峰值信噪比盲估计[J].计算机工程,2017,43(8):253-257.
[18]徐少平,杨荣昌,刘小平.信息量加权的梯度显著度图像质量评价[J].中国图象图形学报,2014,19(2):201-210.