Efficient single image super-resolution via graph-constrained least squares regression

We explore in this paper an efficient algorithmic solution to single image super-resolution (SR). We propose the gCLSR, namely graph-Constrained Least Squares Regression, to super-resolve a high-resolution (HR) image from a single low-resolution (LR) observation. The basic idea of gCLSR is to learn a projection matrix mapping the LR image patch to the HR image patch space while preserving the intrinsic geometric structure of the original HR image patch manifold. Even if gCLSR resembles other manifold learning-based SR methods in preserving the local geometric structure of HR and LR image patch manifolds, the innovation of gCLSR lies in that it preserves the intrinsic geometric structure of the original HR image patch manifold rather than the LR image patch manifold, which may be contaminated by image degeneration (e.g., blurring, down-sampling and noise). Upon acquiring the projection matrix, the target HR image can be simply super-resolved from a single LR image without the need of HR-LR training pairs, which favors resource-limited applications. Experiments on images from the public database show that gCLSR method can achieve competitive quality as state-of-the-art methods, while gCLSR is much more efficient in computation than some state-of-the-art methods.     

基于张量的单幅图像超分辨算法

图像的边缘细节信息直接影响图像的视觉质量。为了尽可能地保持图像边缘结构信息,提高超分辨率图像的质量,提出了一种基于张量的单幅图像超分辨算法。该方法利用张量对图像局部几何特征进行描述,然后根据采样点的局部特征估计待插值点的局部特征,最后通过这一估计的特征计算待插值点的灰度值。实验结果表明基于张量的超分辨方法能够较好地保持图像中的边缘结构信息,峰值信噪比(PSNR)、结构相似性系数(SSIM)等客观评价指标和主观视觉效果都比较好。     

Multipath feedforward network for single image super-resolution

Multimedia Tools and Applications - Single image super-resolution (SR) models which based on convolutional neural network mostly use chained stacking to build the network. It ignores the role of...     

Lightweight refined networks for single image super-resolution

Multimedia Tools and Applications - Recently, software architectures applied to physical agents have become a boost from the emerging Artificial Intelligence(AI). In these smart physical agents,...     

增强的单幅图像自学习超分辨方法

针对图像超分辨率方法构建图像块的稀疏表示（SR）系数存在的主要问题,利用加权思想提出一种增强的单幅图像自学习超分辨方法。首先,通过自学习建立高低分辨率图像金字塔;然后,分别提取低分辨率图像的图像块特征和对应高分辨率图像块的中心像素,并给图像块中不同像素点赋予不同的权重,强调中心像素点在构建图像块稀疏系数时的作用;最后,结合SR理论和支持向量回归（SVR）技术建立超分辨率图像重建模型。实验结果表明,与单幅图像自学习超分辨率方法（SLSR）相比,所提方法的峰值信噪比（PSNR）平均提高了0.39 dB,无参考图像质量评价标准（BRISQUE）分数平均降低了9.7。从主观视角和客观数值证明了所提超分辨率方法更有效。     

Bidirectionally aligned sparse representation for single image super-resolution

It has been demonstrated that the sparse representation based framework is one of the most popular and promising ways to handle the single image super-resolution (SISR) issue. However, due to the complexity of image degradation and inevitable existence of noise, the coding coefficients produced by imposing sparse prior only are not precise enough for faithful reconstructions. In order to overcome it, we present an improved SISR reconstruction method based on the proposed bidirectionally aligned sparse representation (BASR) model. In our model, the bidirectional similarities are first modeled and constructed to form a complementary pair of regularization terms. The raw sparse coefficients are additionally aligned to this pair of standards to restrain sparse coding noise and therefore result in better recoveries. On the basis of fast iterative shrinkage-thresholding algorithm, a well-designed mathematic implementation is introduced for solving the proposed BASR model efficiently. Thorough experimental results indicate that the proposed method performs effectively and efficiently, and outperforms many recently published baselines in terms of both objective evaluation and visual fidelity.     

类别信息生成式对抗网络的单图超分辨重建

目的 基于生成式对抗网络的超分辨模型（SRGAN）以感知损失函数作为优化目标,有效解决了传统基于均方误差（MSE）的损失函数导致重建图像模糊的问题。但是SRGAN的感知损失函数中并未添加明确指示模型生成对应特征的标志性信息,使得其无法精准地将数据的具体维度与语义特征对应起来,受此局限性影响,模型对于生成图像的特征信息表示不足,导致重建结果特征不明显,给后续识别处理过程带来困难。针对上述问题,在SRGAN方法的基础上,提出一种类别信息生成式对抗网络的超分辨模型（class-info SRGAN）。方法 对SRGAN模型增设类别分类器,并将类别损失项添加至生成网络损失中,再利用反向传播训练更新网络参数权重,以达到为模型提供特征类别信息的目的,最终生成具有可识别特征的重建图像。创新及优势在于将特征类别信息引入损失函数,改进了超分辨模型的优化目标,使得重建结果的特征表示更加突出。结果 经CelebA数据集测试表明：添加性别分类器的class-info SRGAN的生成图像性别特征识别率整体偏高（58%97%）;添加眼镜分类器的class-info SRGAN的生成图像眼镜框架更加清晰。此外,模型在Fashion-mnist与Cifar-10数据集上的结果同样表明其相较于SRGAN的重建质量更佳。结论 实验结果验证了本方法在超分辨重建任务中的优势和有效性,同时结果显示：虽然class-info SRGAN更适用于具有简单、具体属性特征的图像,但总体而言仍是一种效果显著的超分辨模型。     

Adaptive deep residual network for single image super-resolution

Computational Visual Media - In recent years, deep learning has achieved great success in the field of image processing. In the single image super-resolution (SISR) task, the convolutional neural...     

Efficient residual attention network for single image super-resolution

Applied Intelligence - The use of deep convolutional neural networks (CNNs) for image super-resolution (SR) from low-resolution (LR) input has achieved remarkable reconstruction performance with...     

Wide receptive field networks for single image super-resolution

Multimedia Tools and Applications - Recently, using deep learning(DL) in super-resolution(SR) has ac- hieved great success. These methods combine the convolutional neural network(CNN) to learn a...     

A novel perceptual loss function for single image super-resolution

In recent years, various deep neural networks have been proposed to improve the performance in the single image super-resolution (SISR) task. The commonly used per-pixel MSE loss function captures less perceptual difference and tends to make the super-resolved images overly smooth, while the perceptual loss function defined on image features extracted from one or two layers of a pretrained network yields more visually pleasing results. We propose a new perceptual loss function via combining features from multiple levels, which incorporates the discrepancy between the reconstruction and the ground truth in different structures. In addition, some variants of the proposed perceptual loss are explored. Extensive quantitative and qualitative comparisons with the state-of-the-art methods demonstrate that our loss function can drive the same network to produce better results when used alone or combined with other loss functions.

     

混合阶通道注意力网络的单图像超分辨率重建

目前用于图像超分辨率重建的通道注意力机制存在注意力预测破坏每个通道和其权重的直接对应关系以及仅仅只考虑一阶或二阶通道注意力而没有综合考虑优势互补的问题，因此提出一种混合阶通道注意力网络的单图像超分辨率重建算法。首先，该网络框架利用局部跨通道相互作用策略将之前一、二阶通道注意力模型采用的升降维改为核为k的一维卷积。这样不仅使得通道注意力预测更直接准确，而且得到的模型相比之前的通道注意力模型更简单；同时，采用改进一、二阶通道注意力模型以综合利用不同阶通道注意力的优势，提高网络判别能力。在基准数据集上的实验结果表明，和现有的超分辨率算法相比，所提算法重建图像的纹理细节和高频信息能得到更好的恢复，且在Set5和BSD100数据集上感知指数（PI）分别平均提高0.3和0.1。这表明此网络能更准确地预测通道注意力并综合利用了不同阶通道注意力，一定程度上提升了性能。     

混合阶通道注意力网络的单图像超分辨率重建

目前用于图像超分辨率重建的通道注意力机制存在注意力预测破坏每个通道和其权重的直接对应关系以及仅仅只考虑一阶或二阶通道注意力而没有综合考虑优势互补的问题,因此提出一种混合阶通道注意力网络的单图像超分辨率重建算法。首先,该网络框架利用局部跨通道相互作用策略将之前一、二阶通道注意力模型采用的升降维改为核为k的一维卷积。这样不仅使得通道注意力预测更直接准确,而且得到的模型相比之前的通道注意力模型更简单;同时,采用改进一、二阶通道注意力模型以综合利用不同阶通道注意力的优势,提高网络判别能力。在基准数据集上的实验结果表明,和现有的超分辨率算法相比,所提算法重建图像的纹理细节和高频信息能得到更好的恢复,且在Set5和BSD100数据集上感知指数（PI）分别平均提高0.3和0.1。这表明此网络能更准确地预测通道注意力并综合利用了不同阶通道注意力,一定程度上提升了性能。     

用于单幅模糊图像超分辨的Transformer融合网络

目的 以卷积神经网络为代表的深度学习方法已经在单帧图像超分辨领域取得了丰硕成果,这些方法大多假设低分辨图像不存在模糊效应。然而,由于相机抖动、物体运动等原因,真实场景下的低分辨率图像通常会伴随着模糊现象。因此,为了解决模糊图像的超分辨问题,提出了一种新颖的Transformer融合网络。方法 首先使用去模糊模块和细节纹理特征提取模块分别提取清晰边缘轮廓特征和细节纹理特征。然后,通过多头自注意力机制计算特征图任一局部信息对于全局信息的响应,从而使Transformer融合模块对边缘特征和纹理特征进行全局语义级的特征融合。最后,通过一个高清图像重建模块将融合特征恢复成高分辨率图像。结果 实验在2个公开数据集上与最新的9种方法进行了比较,在GOPRO数据集上进行2倍、4倍、8倍超分辨重建,相比于性能第2的模型GFN(gated fusion network),峰值信噪比(peak signal-to-noive ratio,PSNR)分别提高了0.12 d B、0.18 d B、0.07 d B;在Kohler数据集上进行2倍、4倍、8倍超分辨重建,相比于性能第2的模型GFN,PSNR值分别...     

Wavelet integrated residual dictionary training for single image super-resolution

Multimedia Tools and Applications - Single image super-resolution (SISR) is an encouraging process for image processing applications, where the super-resolution (SR) problem is resolved by the...     

基于级联深度卷积神经网络的高性能图像超分辨率重构

为了进一步提高现有图像超分辨率重构方法所得图像的分辨率,提出一种高性能的深度卷积神经网络（HDCN）模型用于重构放大倍数固定的超分辨率图像。通过建立级联HDCN模型解决传统模型重构图像时放大倍数无法按需选择的问题,并在级联过程中引入深度边缘滤波器以减少级联误差,突出边缘信息,从而得到高性能的级联深度卷积神经网络（HCDCN）模型。基于Set5、Set14数据集进行超分辨率图像重构实验,证明了引入深度边缘滤波器的有效性,对比HCDCN方法与其他图像超分辨率重构方法的性能评估结果,展现了HCDCN方法的优越性能。     

Learning an epipolar shift compensation for light field image super-resolution

Light field imaging has drawn broad attention since the advent of practical light field capturing systems that facilitate a wide range of applications in computer vision. However, existing learning-based methods for improving the spatial resolution of light field images neglect the shifts in the sub-pixel domain that are widely used by super-resolution techniques, thus, fail in recovering rich high-frequency information. To fully exploit the shift information, our method attempts to learn an epipolar shift compensation for light field image super-resolution that allows the restored light field image to be angular coherent with the enhancement of spatial resolution. The proposed method first utilizes the rich surrounding views along some typical epipolar directions to explore the inter-view correlations. We then implement feature-level registration to capture accurate sub-pixel shifts of central view, which is constructed by the compensation module equipped with dynamic deformable convolution. Finally, the complementary information from different spatial directions is fused to provide high-frequency details for the target view. By taking each sub-aperture image as a central view, our method could be applied for light field images with any angular resolution. Extensive experiments on both synthetic and real scene datasets demonstrate the superiority of our method over the state-of-the-art qualitatively and quantitatively. Moreover, the proposed method shows good performance in preserving the inherent epipolar structures in light field images. Specifically, our LFESCN method outperforms the state-of-the-art method with about 0.7 dB (PSNR) on average.     

基于结构相似度的单幅图像正则超分辨复原

图像具有大量的局部结构相似区域,并且这种相似性可以在多个尺度上保持。基于这一特征,利用结构相似指标进行相似性匹配生成相似的低分辨率图像序列,从而把单幅图像的超分辨问题转化为图像序列超分辨问题来解决。文中提出了一种新的自适应的正则化方法,正则参数的选取使得目标函数存在全局最优解。最后证明了算法的收敛性。实验表明,该方法具有很好的复原效果。     

Clustering based multiple branches deep networks for single image super-resolution

Since the limitation of optical sensors, it’s often hard to obtain an image with the ideal resolution. Image super-resolution (SR) technology can generate a high-resolution image from the corresponding low-resolution image. Recently, deep learning (DL) based SR methods draw much attention due to their satisfying reconstruction results. However, these methods often neglect the diversity of image patches. Therefore, the reconstruction effect is limited. To fully exploit the texture variability across different image patches, we propose a universal, flexible, and effective framework. The proposed framework can be adopted to any DL based methods. It can significantly improve the SR accuracy while maintaining the running time. In the proposed framework, K-means is employed to cluster image patches into different categories. Multiple CNN branches are designed for these different categories to reconstruct the SR image. Each branch is weighted in accordance with the Euclidean distance to the cluster centers. Experimental results demonstrate that by applying the proposed framework, performance of the DL based SR method can be significantly improved.