共查询到20条相似文献，搜索用时 46 毫秒
1.
图像分辨率是衡量遥感图像质量的重要指标,受限于成像设备和传输条件,传统遥感图像的清晰度难以保证,针对上述问题,提出了一种基于条件生成对抗网络的遥感图像超分辨率重建的改进模型。为了加快模型的收敛速度,在生成器网络中使用内容损失和对抗损失相结合作为目标函数。另外为了提高了网络训练的稳定性,在判别器网络中引入梯度惩罚函数对判别器梯度进行限制。实验结果表明,改进后的模型相较于SRCNN、FSRCNN和SRGAN模型,主观视觉效果和客观评价指标均有显著提升。 相似文献
2.
3.
针对现有图像超分辨率重建方法中高频图像信息不丰富的问题,提出一种基于反馈和注意机制的单图像重建生成对抗网络(GFSRGAN)。采用反馈网络作为生成器,通过反馈连接逐步生成高分辨率图像;提出一种具有注意机制的反馈块,其能在处理反馈流的同时,自适应地选择有用的特征信息;利用相对平均最小二乘GAN(Ra LSGAN)损失引导模型获得更真实的图像。实验结果表明,与现有基于GAN的超分辨方法相比,该方法重建出的图像纹理更加逼真自然。 相似文献
4.
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。 相似文献
5.
6.
本文针对现有光学遥感图像超分辨率重建模型对感受野尺度关注不足和对特征通道信息提取不充分带来的问题, 提出了一种基于多尺度特征提取和坐标注意力的光学遥感图像超分辨率重建模型. 该重建模型基于深度残差网络结构, 在网络的高频分支中设计了多个级联的多尺度特征和坐标注意力模块 (multi-scale feature & coordinate attention block, MFCAB), 对输入的低分辨率光学遥感图像的高频特征进行充分发掘: 首先, 在MFCAB模块中引入Inception子模块, 使用不同尺度的卷积核捕捉不同感受野下的空间特征; 其次, 在Inception子模块后增加坐标注意力子模块, 同时关注通道与坐标两个维度, 以获得更好的通道注意力效果; 最后, 对各MFCAB模块提取的特征进行多路径融合, 实现多重多尺度空间信息与通道注意信息的有效融合. 本文模型在NWPU4500数据集上2倍、3倍放大中PSNR值达到34.73 dB和30.12 dB, 较EDSR分别提升0.66 dB和0.01 dB, 在AID1600数据集上2倍、3倍、4倍放大中PSNR值达到34.71 dB、30.58 dB、28.44 dB, 较EDSR分别提升0.09 dB、0.03 dB、0.04 dB. 实验结果表明, 该模型在光学遥感图像数据集上的重建效果优于主流的图像超分辨率重建模型. 相似文献
7.
受井下粉尘大、照度低等环境影响,矿井图像存在分辨率低、细节模糊等问题,现有的图像超分辨率重建算法应用于矿井图像时,难以获取不同尺度图像信息、网络参数过大而影响重建速度,且重建图像易出现细节丢失、边缘轮廓模糊、伪影等问题。提出了一种基于多尺度密集通道注意力超分辨率生成对抗网络（SRGAN）的矿井图像超分辨率重建算法。设计了多尺度密集通道注意力残差块替代SRGAN原有的残差块,采用2路并行且卷积核大小不同的密集连接块,可充分获取图像特征;融入高效通道注意力模块,加强对高频信息的关注度;采用深度可分离卷积对网络进行轻量化,抑制网络参数的增加;利用纹理损失约束网络训练,避免网络加深时产生伪影。在井下数据集和公共数据集上对提出的矿井图像超分辨率重建算法和经典超分辨率重建算法BICUBIC,SRCNN,SRRESNET,SRGAN进行实验,结果表明：所提算法在主客观评价上总体优于对比算法,网络参数较SRGAN减少了2.54%,峰值信噪比与结构相似度较经典算法指标均值分别提高了0.764 dB和0.053 58,能更好地关注图像的纹理、轮廓等细节信息,重建图像更符合人眼视觉。 相似文献
8.
受井下粉尘大、照度低等环境影响,矿井图像存在分辨率低、细节模糊等问题,现有的图像超分辨率重建算法应用于矿井图像时,难以获取不同尺度图像信息、网络参数过大而影响重建速度,且重建图像易出现细节丢失、边缘轮廓模糊、伪影等问题。提出了一种基于多尺度密集通道注意力超分辨率生成对抗网络（SRGAN）的矿井图像超分辨率重建算法。设计了多尺度密集通道注意力残差块替代SRGAN原有的残差块,采用2路并行且卷积核大小不同的密集连接块,可充分获取图像特征;融入高效通道注意力模块,加强对高频信息的关注度;采用深度可分离卷积对网络进行轻量化,抑制网络参数的增加;利用纹理损失约束网络训练,避免网络加深时产生伪影。在井下数据集和公共数据集上对提出的矿井图像超分辨率重建算法和经典超分辨率重建算法BICUBIC,SRCNN,SRRESNET,SRGAN进行实验,结果表明：所提算法在主客观评价上总体优于对比算法,网络参数较SRGAN减少了2.54%,峰值信噪比与结构相似度较经典算法指标均值分别提高了0.764 dB和0.053 58,能更好地关注图像的纹理、轮廓等细节信息,重建图像更符合人眼视觉。 相似文献
9.
现有的基于卷积神经网络的超分辨率重建方法由于感受野限制, 难以充分利用遥感图像丰富的上下文信息和自相关性, 导致重建效果不佳. 针对该问题, 本文提出了一种基于多重蒸馏与Transformer的遥感图像超分辨率(remote sensing image super-resolution based on multi-distillation and Transformer, MDT)重建方法. 首先结合多重蒸馏和双注意力机制, 逐步提取低分辨率图像中的多尺度特征, 以减少特征丢失. 接着, 构建一种卷积调制Transformer来提取图像的全局信息, 恢复更多复杂的纹理细节, 从而提升重建图像的视觉效果. 最后, 在上采样过程中添加全局残差路径, 提高特征在网络中的传播效率, 有效减少了图像的失真与伪影问题. 在AID和UCMerced两个数据集上的进行实验, 结果表明, 本文方法在放大至4倍超分辨率任务上的峰值信噪比和结构相似度分别最高达到了29.10 dB和0.7807, 重建图像质量明显提高, 并且在细节保留方面达到了更好的视觉效果. 相似文献
10.
传统生成对抗网络模型重建人脸图像时出现过多失真,难以在减少失真的情况下有效提高人脸图像真实感.针对该问题,在生成对抗网络SRGAN模型的基础上,提出一种改进的人脸图像超分辨率重建方法.为提高重建像素点与周围像素点的相关性,将双注意力机制模块嵌入到SRGAN模型的生成器和判别器中,在空间域和通道域中获取更精准的特征依赖关系.同时应用自适应激活函数ACON取代原SRGAN网络中的激活函数,通过动态学习ACON激活函数参数为每个神经元设计不同激活形式,从而提高网络特征表达能力.使用改进SRGAN的人脸图像超分辨率重建算法在CelebA测试集上进行重建实验,结果表明:该算法较原算法PSNR值提高0.675 dB,SSIM值提高0.016,LPIPS值优化0.036,有效减少了重建人脸图像中眼睛等重点部位的失真情况;与其他非生成对抗网络的主流算法相比,LPIPS值最低优化0.107,最高优化0.205,有效提高了重建人脸图像的真实感. 相似文献
11.
12.
本次研究主要源自于单张图像中的雨滴去除问题。由于附着在玻璃窗或者相机镜头上的雨滴会严重降低图像的质量,因而使用生成对抗网络将带有雨滴的退化图像转换为干净的图像来解决此问题。文章的主要思想是将视觉注意力机制引入到生成网络中,提出新的残差U-Nets处理注意力图,并使用判别网络进行甄别图像的真实性,同时使用新的感知损失作为损失函数,从而使网络可以更加关注雨滴区域及其周围的环境。这样处理不但可以使恢复的图像具有更高的质量,同时也能具有更加优秀的视觉效果。文章采用峰值信噪比和结构相似性作为模型的数值评价标准,图像的细节展示作为视觉评价标准。实验表明,这种方法无论在视觉效果,还是数值结果上都具有不错的表现。 相似文献
13.
14.
遥感图像的空间分辨率高,不同类型对象的尺度差异大、类别不平衡,是精准语义分割任务所面临的主要挑战。为了提高遥感图像语义分割的准确性,提出了一种改进U-Net的多尺度特征融合遥感图像语义分割网络(Multi-scale Feature Fusion Network,MFFNet)。该网络以U-Net网络为基础,包含动态特征融合模块和门控注意力卷积混合模块。其中,动态特征融合模块代替跳跃连接,改进上采样层和下采样层的特征融合方式,避免特征融合导致信息丢失,同时提高浅层特征和深层特征的融合效果;门控注意力卷积混合模块通过整合自注意力、卷积和门控机制,更好地捕获局部和全局信息。在Potsdam和Vaihingen数据集上开展对比实验和消融实验,结果表明MFFNet在两个数据集上的mIoU分别达到76.95%和72.93%,有效提高了遥感图像的语义分割精度。 相似文献
15.
《Displays》2023
It is a great challenge to rendering glare on image as the current rendering algorithms did not consider well the refraction of human eyes, thus the effect of rendering, in some critical application such as vehicle headlamps, is not real and may affect the safety evaluation. The traditional glare rendering algorithm relies on a large number of hand-designed wave optics processing operators, not only cannot complete the rendering work online in real time, but also cannot cope with the complex and changeable imaging conditions in reality. The mainstream generative adversarial network based algorithms in the field of image style translation are introduced to generate glare effect, which could be rendering online in a real time, however they still fail to render some effects such as detail distortion. In this work, we present a novel glare simulation generation method which is the first algorithm to apply a generative model based style transfer method to glare rendering. In a nutshell, a new method named Glare Generation Network is proposed to aggregate the benefits of content diversity and style consistency, which combines both paired and unpaired branch in a dual generative adversarial network. Our approach increase the structural similarity index measure by at least 0.039 on the custom darkroom vehicle headlamp dataset. We further show our method significantly improve the inference speed. 相似文献
16.
《Displays》2021
Arterial spin labeling (ASL) is a relatively new MRI technique that can measure cerebral blood flow, which is of great importance for the diagnosis of dementia diseases. Besides, this valuable imaging modality does not need exogenous tracers and has no radiation, which makes it favorable for elder patients. However, ASL data does lack in many contemporary image-based dementia diseases datasets, which include popular ADNI-1/GO/2/3 datasets. In order to supplement the valuable ASL data, a new Generative adversarial network (GAN)-based model is proposed to synthesize ASL images in this study. This new model is unique, as the popular variational auto-encoder (VAE) has been utilized as the generator of the GAN-based model. Hence, a new VAE-GAN architecture is introduced in this study. In order to demonstrate its superiority, dozens of experiments have been conducted. Experimental results demonstrate that, this new VAE-GAN model is superior to other state-of-the-art ASL image synthesis methods, and the accuracy improvement after incorporating synthesized ASL images from the new model can be as high as 42.41% in dementia diagnosis tasks. 相似文献
17.
Oceanic eddies and fronts are plentiful in the global oceans. There are abundant frontal processes at the edges of eddies, meanwhile, eddies can be formed due to instability along the fronts. Moreover, they can both induce submesoscale processes and strong vertical motions. Consequently, it is crucial to understand the correlation between eddies and fronts, along with their spatiotemporal distribution characteristics, for the transport of ocean energy and material and marine ecosystems. Generally, using remote sensing to detect eddies and fronts is based on geometrics, physical parameters, and handcrafted features. Existing approaches are inaccurate due to the highly dynamic nature of eddies and fronts as well as the lack of physical mechanisms between them. This paper proposes an adversarial approach for mining eddy-front coupling, dubbed the eddy-front generative adversarial network (EFGAN). In EFGAN, the generator follows the encoder–decoder structure and consists of a data encoder, a feature decoder, and a multi-task generation module. The rich contextual and semantic information on eddy features and frontal structure, which are the physical constraints for EFGAN, can be extracted from the fusion of satellite sea level anomaly (SLA) and sea surface temperature (SST) data, generating styled eddy-front coupling images with the accompaniment of the mask of eddy-front categories and styled fronts. To tackle the issue of inconsistency in the category of a single eddy, an instance consistency loss is designed to make the category of each individual eddy as consistent as possible, thus ensuring that the correlation of eddy-front belongs to a single category. Furthermore, multiple loss functions are combined to jointly guide network training to improve stability. Extensive experiments demonstrate that EFGAN outperforms state-of-the-art generative adversarial network-based models. The spatial distribution indicates that frontal eddies are more prevalent at the confluence where warm and cold currents meet. In the Kuroshio Extension region, there are concentrated distribution areas of frontal eddies that are strong in the boreal spring and summer, while eddy-induced fronts are highly active in the boreal winter and spring but weak in autumn. 相似文献
18.
This study proposes a unified gradient- and intensity-discriminator generative adversarial network for various image fusion tasks, including infrared and visible image fusion, medical image fusion, multi-focus image fusion, and multi-exposure image fusion. On the one hand, we unify all fusion tasks into discriminating a fused image’s gradient and intensity distributions based on a generative adversarial network. The generator adopts a dual-encoder–single-decoder framework to extract source image features by using different encoder paths. A dual-discriminator is employed to distinguish the gradient and intensity, ensuring that the generated image contains the desired geometric structure and conspicuous information. The dual adversarial game can tackle the generative adversarial network’s mode collapse problem. On the other hand, we define a loss function based on the gradient and intensity that can be adapted to various fusion tasks by using varying relevant parameters with the source images. Qualitative and quantitative experiments on publicly available datasets demonstrate our method’s superiority over state-of-the-art methods. 相似文献
19.