针对当前电力设备红外图像分辨率低和温度分布模糊问题,提出一种基于局部和全局信息注意力生成对抗网络(local and global information attention generative adversarial network,LGIA-GAN)的超分辨率重建方法。首先,使用门控权重单元...针对当前电力设备红外图像分辨率低和温度分布模糊问题,提出一种基于局部和全局信息注意力生成对抗网络(local and global information attention generative adversarial network,LGIA-GAN)的超分辨率重建方法。首先,使用门控权重单元融合多种卷积输出构建细节增强融合卷积,增加重要信息在输出特征图的占比;其次,搭建双注意力模块,对图像长距离像素依赖关系建模并捕获空间和通道维度信息;然后,构造生成对抗网络,使网络关注电力设备红外图像局部纹理细节和全局轮廓信息;最后,通过实验证明,LGIA-GAN在数据集上的峰值信噪比和结构相似度分别为30.266dB和0.9197,重建时间为0.120s,明显优于其他几种GAN算法,并在主观视觉上重建效果更好。所提方法能够有效提升电力设备热成像分辨率,对电力设备故障诊断具有支撑作用。展开更多
针对现有遥感图像超分辨率重建算法存在模型复杂度高、多尺度特征利用不充分等问题,提出一种用于遥感图像超分辨率重建的轻量多尺度特征融合网络(RSMFFN)。首先,利用部分卷积PConv设计了一种轻量的特征提取块LEB,结合LEB与增强空间注意...针对现有遥感图像超分辨率重建算法存在模型复杂度高、多尺度特征利用不充分等问题,提出一种用于遥感图像超分辨率重建的轻量多尺度特征融合网络(RSMFFN)。首先,利用部分卷积PConv设计了一种轻量的特征提取块LEB,结合LEB与增强空间注意力设计轻量多尺度特征融合块LMFB,自适应捕获不同尺度特征间的语义关联并对特征进行初步融合;其次,设计了基于高效通道注意力的层级特征融合机制HFF用于跨层特征融合,利用浅层高频信息指导深层特征重建,实现多级特征的协同优化。最后,将多个多尺度特征融合块LMFB堆叠组成特征蒸馏组FDG,并利用HFF进行特征融合实现特征蒸馏,在提取深层有效特征的同时逐步对特征进行细化。在UCMerced LandUse data set数据集上的实验表明,所提出的网络参数量及计算量都显著减少,峰值信噪比以及结构相似度均有所提升,在保持轻量化优势的同时,也保证了遥感图像的超分辨率重建质量。展开更多
In our study, support vector value contourlet transform is constructed by using support vector regression model and directional filter banks. The transform is then used to decompose source images at multi-scale, multi...In our study, support vector value contourlet transform is constructed by using support vector regression model and directional filter banks. The transform is then used to decompose source images at multi-scale, multi-direction and multi-resolution. After that, the super-resolved multi-spectral image is reconstructed by utilizing the strong learning ability of support vector regression and the correlation between multi-spectral image and panchromatic image. Finally, the super-resolved multi- spectral image and the panchromatic image are fused based on regions at different levels. Our experi- ments show that, the learning method based on support vector regression can improve the effect of super-resolution of multi-spectral image. The fused image preserves both high space resolution and spectrum information of multi-spectral image.展开更多
A multi-channel fast super-resolution image reconstruction algorithm based on matrix observation model is proposed in the paper,which consists of three steps to avoid the computational complexity: a single image SR re...A multi-channel fast super-resolution image reconstruction algorithm based on matrix observation model is proposed in the paper,which consists of three steps to avoid the computational complexity: a single image SR reconstruction step,a registration step and a wavelet-based image fusion. This algorithm decomposes two large matrixes to the tensor product of two little matrixes and uses the natural isomorphism between matrix space and vector space to transform cost function based on matrix-vector products model to matrix form. Furthermore,we prove that the regularization part can be transformed to the matrix formed. The conjugate-gradient method is used to solve this new model. Finally,the wavelet fusion is used to integrate all the registered highresolution images obtained from the single image SR reconstruction step. The proposed algorithm reduces the storage requirement and the calculating complexity,and can be applied to large-dimension low-resolution images.展开更多
文摘针对当前电力设备红外图像分辨率低和温度分布模糊问题,提出一种基于局部和全局信息注意力生成对抗网络(local and global information attention generative adversarial network,LGIA-GAN)的超分辨率重建方法。首先,使用门控权重单元融合多种卷积输出构建细节增强融合卷积,增加重要信息在输出特征图的占比;其次,搭建双注意力模块,对图像长距离像素依赖关系建模并捕获空间和通道维度信息;然后,构造生成对抗网络,使网络关注电力设备红外图像局部纹理细节和全局轮廓信息;最后,通过实验证明,LGIA-GAN在数据集上的峰值信噪比和结构相似度分别为30.266dB和0.9197,重建时间为0.120s,明显优于其他几种GAN算法,并在主观视觉上重建效果更好。所提方法能够有效提升电力设备热成像分辨率,对电力设备故障诊断具有支撑作用。
文摘针对现有遥感图像超分辨率重建算法存在模型复杂度高、多尺度特征利用不充分等问题,提出一种用于遥感图像超分辨率重建的轻量多尺度特征融合网络(RSMFFN)。首先,利用部分卷积PConv设计了一种轻量的特征提取块LEB,结合LEB与增强空间注意力设计轻量多尺度特征融合块LMFB,自适应捕获不同尺度特征间的语义关联并对特征进行初步融合;其次,设计了基于高效通道注意力的层级特征融合机制HFF用于跨层特征融合,利用浅层高频信息指导深层特征重建,实现多级特征的协同优化。最后,将多个多尺度特征融合块LMFB堆叠组成特征蒸馏组FDG,并利用HFF进行特征融合实现特征蒸馏,在提取深层有效特征的同时逐步对特征进行细化。在UCMerced LandUse data set数据集上的实验表明,所提出的网络参数量及计算量都显著减少,峰值信噪比以及结构相似度均有所提升,在保持轻量化优势的同时,也保证了遥感图像的超分辨率重建质量。
基金Supported by the National Natural Science Foundation of China(61172127)Key Research Project of Education Department of Anhui Province(KJ2010A021)
文摘In our study, support vector value contourlet transform is constructed by using support vector regression model and directional filter banks. The transform is then used to decompose source images at multi-scale, multi-direction and multi-resolution. After that, the super-resolved multi-spectral image is reconstructed by utilizing the strong learning ability of support vector regression and the correlation between multi-spectral image and panchromatic image. Finally, the super-resolved multi- spectral image and the panchromatic image are fused based on regions at different levels. Our experi- ments show that, the learning method based on support vector regression can improve the effect of super-resolution of multi-spectral image. The fused image preserves both high space resolution and spectrum information of multi-spectral image.
基金Sponsored by the National Natural Science Foundation of China(Grant No.60474016)the Fundamental Research Funds for the Central Universities(Grant No.HIT.NSRIF.2009046)
文摘A multi-channel fast super-resolution image reconstruction algorithm based on matrix observation model is proposed in the paper,which consists of three steps to avoid the computational complexity: a single image SR reconstruction step,a registration step and a wavelet-based image fusion. This algorithm decomposes two large matrixes to the tensor product of two little matrixes and uses the natural isomorphism between matrix space and vector space to transform cost function based on matrix-vector products model to matrix form. Furthermore,we prove that the regularization part can be transformed to the matrix formed. The conjugate-gradient method is used to solve this new model. Finally,the wavelet fusion is used to integrate all the registered highresolution images obtained from the single image SR reconstruction step. The proposed algorithm reduces the storage requirement and the calculating complexity,and can be applied to large-dimension low-resolution images.
