由于煤矿井下特殊的环境导致井下图像一般比较模糊和边缘特征比较弱,而目前基于字典学习的图像超分辨率重建算法通常是将全部图像块利用训练的单一字典进行重建,忽视了各图像块之间的差异性,不利于重建边缘不清晰的矿井图像。结合矿井...由于煤矿井下特殊的环境导致井下图像一般比较模糊和边缘特征比较弱,而目前基于字典学习的图像超分辨率重建算法通常是将全部图像块利用训练的单一字典进行重建,忽视了各图像块之间的差异性,不利于重建边缘不清晰的矿井图像。结合矿井图像特征提出一种边缘融合的多字典超分辨率图像重建方法,该算法根据各图像块的梯度统计信息将图像块进行分类并训练对应的字典库,重建时将不同字典重建的图像块融合成完整的高分辨率图像;此外为了提高图像的边缘信息,预处理阶段低分辨率图像进行边缘融合以增强边缘特征,重建的高分辨率图像利用学习的先验知识进行边缘融合以修正重建过程中出现的误差。实验表明,该算法的重建效果优于其它基于字典学习的超分辨率图像重建方法,能够很好地重建图像的边缘细节,并抑制重建过程中产生的重影和振铃效应,平均PSNR值提高1.19 d B。展开更多
针对模型VDSR(very deep super resolution)收敛速度慢,训练前需要对原始图像进行预处理,以及网络中存在的冗余性等问题,提出了一种基于深度跳跃级联的单幅图像超分辨率重建(DCSR)算法。DCSR算法省去了图像预处理,直接在低分辨率图像上...针对模型VDSR(very deep super resolution)收敛速度慢,训练前需要对原始图像进行预处理,以及网络中存在的冗余性等问题,提出了一种基于深度跳跃级联的单幅图像超分辨率重建(DCSR)算法。DCSR算法省去了图像预处理,直接在低分辨率图像上提取浅层特征,并使用亚像素卷积对图像进行放大;通过使用跳跃级联块可以充分利用每个卷积层提取到图像特征,实现特征重用,减少网络的冗余性。网络的跳跃级联块可以直接从输出到每一层建立短连接,加快网络的收敛速度,缓解梯度消失问题。实验结果表明,在几种公开数据集上,所提算法的峰值信噪比、结构相似度值均高于现有的几种算法,充分证明了所提算法的出色性能。展开更多
An image magnification method with a Gradient Vector Flow(GVF)constraint-basedanisotropic diffusion model is proposed in this letter.A Low-Resolution(LR)image is first magnifiedusing bilinear interpolation,and then an...An image magnification method with a Gradient Vector Flow(GVF)constraint-basedanisotropic diffusion model is proposed in this letter.A Low-Resolution(LR)image is first magnifiedusing bilinear interpolation,and then an iterative image restoration method,with the use of an ani-sotropic diffusion model and a Gaussian moving-average constraint,is applied to the magnified image.The estimated GVF of a High-Resolution(HR)image can be used to remove the jagged effect and topreserve the textural structure in the image.Meanwhile,the use of the Gaussian moving-average LRmodel can provide a data fidelity constraint,which renders a magnified image closer to the ideal HRversion.Experimental results show that the proposed method can improve the quality of magnifiedimages in terms of both objective and subjective criteria.展开更多
This paper designs a 3 mm radiometer and validate with experiments based on the principle of passive millimeter wave （PMMW） imaging. The poor spatial resolution, which is limited by antenna size, should be improved ...This paper designs a 3 mm radiometer and validate with experiments based on the principle of passive millimeter wave （PMMW） imaging. The poor spatial resolution, which is limited by antenna size, should be improved by post data processing. A conjugate-gradient （CG） algorithm is adopted to circumvent this drawback. Simulation and real data collected in laboratory environment are given, and the results show that the CG algorithm improves the spatial resolution and convergent rate. Further, it can reduce the ringing effects which are caused by regularizing the image restoration. Thus, the CG algorithm is easily implemented for PMMW imaging.展开更多
文摘由于煤矿井下特殊的环境导致井下图像一般比较模糊和边缘特征比较弱,而目前基于字典学习的图像超分辨率重建算法通常是将全部图像块利用训练的单一字典进行重建,忽视了各图像块之间的差异性,不利于重建边缘不清晰的矿井图像。结合矿井图像特征提出一种边缘融合的多字典超分辨率图像重建方法,该算法根据各图像块的梯度统计信息将图像块进行分类并训练对应的字典库,重建时将不同字典重建的图像块融合成完整的高分辨率图像;此外为了提高图像的边缘信息,预处理阶段低分辨率图像进行边缘融合以增强边缘特征,重建的高分辨率图像利用学习的先验知识进行边缘融合以修正重建过程中出现的误差。实验表明,该算法的重建效果优于其它基于字典学习的超分辨率图像重建方法,能够很好地重建图像的边缘细节,并抑制重建过程中产生的重影和振铃效应,平均PSNR值提高1.19 d B。
文摘针对模型VDSR(very deep super resolution)收敛速度慢,训练前需要对原始图像进行预处理,以及网络中存在的冗余性等问题,提出了一种基于深度跳跃级联的单幅图像超分辨率重建(DCSR)算法。DCSR算法省去了图像预处理,直接在低分辨率图像上提取浅层特征,并使用亚像素卷积对图像进行放大;通过使用跳跃级联块可以充分利用每个卷积层提取到图像特征,实现特征重用,减少网络的冗余性。网络的跳跃级联块可以直接从输出到每一层建立短连接,加快网络的收敛速度,缓解梯度消失问题。实验结果表明,在几种公开数据集上,所提算法的峰值信噪比、结构相似度值均高于现有的几种算法,充分证明了所提算法的出色性能。
基金a grant from the Research Grants Council othe Hong Kong Special Administrative Region,China(NoPolyU 5199/06E)by the National Natural ScienceFoundation of China(No.60472036,No.60431020,No60402036,No.60772069)the Natural Science Foundation of Beijing(No.4062006).
文摘An image magnification method with a Gradient Vector Flow(GVF)constraint-basedanisotropic diffusion model is proposed in this letter.A Low-Resolution(LR)image is first magnifiedusing bilinear interpolation,and then an iterative image restoration method,with the use of an ani-sotropic diffusion model and a Gaussian moving-average constraint,is applied to the magnified image.The estimated GVF of a High-Resolution(HR)image can be used to remove the jagged effect and topreserve the textural structure in the image.Meanwhile,the use of the Gaussian moving-average LRmodel can provide a data fidelity constraint,which renders a magnified image closer to the ideal HRversion.Experimental results show that the proposed method can improve the quality of magnifiedimages in terms of both objective and subjective criteria.
基金supported partly by the State Key Program of National Natural Science Foundation of China(60632020)the Youth Science Foundation of University of Electronic Science and Technology of China(JX0823).
文摘This paper designs a 3 mm radiometer and validate with experiments based on the principle of passive millimeter wave （PMMW） imaging. The poor spatial resolution, which is limited by antenna size, should be improved by post data processing. A conjugate-gradient （CG） algorithm is adopted to circumvent this drawback. Simulation and real data collected in laboratory environment are given, and the results show that the CG algorithm improves the spatial resolution and convergent rate. Further, it can reduce the ringing effects which are caused by regularizing the image restoration. Thus, the CG algorithm is easily implemented for PMMW imaging.
