贝叶斯网络(bayesian network,BN)小数据集条件下,定性最大后验概率(qualitative maximum a posteriori,QMAP)估计往往会违反给定的专家约束,这就导致QMAP估计偏离真实值。为了克服该算法的缺陷,提出了一种改进的QMAP算法。首先,学习出Q...贝叶斯网络(bayesian network,BN)小数据集条件下,定性最大后验概率(qualitative maximum a posteriori,QMAP)估计往往会违反给定的专家约束,这就导致QMAP估计偏离真实值。为了克服该算法的缺陷,提出了一种改进的QMAP算法。首先,学习出QMAP估计,再结合保序回归方法对违反不等式约束的参数进行校正;然后使用一种微调策略对校正后的参数做进一步调整,使所得参数能够满足专家约束;最后,与最大似然估计(maximum likelihood estimation,MLE)和QMAP算法对比。仿真实验结果表明:在小数据集条件下,提出的算法满足所有约束条件,KL(Kullback-Leibler)散度始终低于其他2种算法,运行时间高于其他2种算法约0.1 s,影响甚微,且推理结果贴近真实值,偏差维持在±0.05之间。改进的QMAP算法的综合性能优于MLE、QMAP算法,并具有较好的实用性。展开更多
A maximum a posteriori（ MAP） algorithm is proposed to improve the accuracy of super resolution（ SR） reconstruction in traditional methods. The algorithm applies both joints image registration and SR reconstruction...A maximum a posteriori（ MAP） algorithm is proposed to improve the accuracy of super resolution（ SR） reconstruction in traditional methods. The algorithm applies both joints image registration and SR reconstruction in the framework,but separates them in the process of iteratiion. Firstly,we estimate the shifting parameters through two lowresolution（ LR） images and use the parameters to reconstruct initial HR images. Then,we update the shifting parameters using HR images. The aforementioned steps are repeated until the ideal HR images are obtained. The metrics such as PSNR and SSIM are used to fully evaluate the quality of the reconstructed image. Experimental results indicate that the proposed method can enhance image resolution efficiently.展开更多
The paper presents an algorithm of automatic target detection in Synthetic Aperture Radar(SAR) images based on Maximum A Posteriori(MAP). The algorithm is divided into three steps. First, it employs Gaussian mixture d...The paper presents an algorithm of automatic target detection in Synthetic Aperture Radar(SAR) images based on Maximum A Posteriori(MAP). The algorithm is divided into three steps. First, it employs Gaussian mixture distribution to approximate and estimate multi-modal histogram of SAR image. Then, based on the principle of MAP, when a priori probability is both unknown and learned respectively, the sample pixels are classified into different classes c = {target,shadow, background}. Last, it compares the results of two different target detections. Simulation results preferably indicate that the presented algorithm is fast and robust, with the learned a priori probability, an approach to target detection is reliable and promising.展开更多
Interferogram noise reduction is a very important processing step in Interferometric Synthetic Aperture Radar(InSAR) technique. The most difficulty for this step is to remove the noises and preserve the fringes simult...Interferogram noise reduction is a very important processing step in Interferometric Synthetic Aperture Radar(InSAR) technique. The most difficulty for this step is to remove the noises and preserve the fringes simultaneously. To solve the dilemma, a new interferogram noise reduction algorithm based on the Maximum A Posteriori(MAP) estimate is introduced in this paper. The algorithm is solved under the Total Generalized Variation(TGV) minimization assumption, which exploits the phase characteristics up to the second order differentiation. The ideal noise-free phase consisting of piecewise smooth areas is involved in this assumption, which is coincident with the natural terrain. In order to overcome the phase wraparound effect, complex plane filter is utilized in this algorithm. The simulation and real data experiments show the algorithm can reduce the noises effectively and meanwhile preserve the interferogram fringes very well.展开更多
In this paper we consider three problems in continuous multi-criteria optimization: An application of the Berge Maximum Theorem, properties of Pareto-retract mappings, and the structure of Pareto sets. The key goal of...In this paper we consider three problems in continuous multi-criteria optimization: An application of the Berge Maximum Theorem, properties of Pareto-retract mappings, and the structure of Pareto sets. The key goal of this work is to present the relationship between the three problems mentioned above. First, applying the Maximum Theorem we construct the Pareto-retract mappings from the feasible domain onto the Pareto-optimal solutions set if the feasible domain is compact. Next, using these mappings we analyze the structure of the Pareto sets. Some basic topological properties of the Pareto solutions sets in the general case and in the convex case are also discussed.展开更多
现有优秀的基于深度学习的分布式视频压缩感知(Distributed Compressed Video Sensing,DCVS)重构算法利用测量值和参考帧顺序更新非关键帧,获得了较好的重构性能,但由于缺乏较严格的理论指导,无法充分结合这两类信息,限制了非关键帧重...现有优秀的基于深度学习的分布式视频压缩感知(Distributed Compressed Video Sensing,DCVS)重构算法利用测量值和参考帧顺序更新非关键帧,获得了较好的重构性能,但由于缺乏较严格的理论指导,无法充分结合这两类信息,限制了非关键帧重构质量的进一步提升.针对该问题,本文首先利用贝叶斯理论及最大后验概率(Maximum A Posteriori,MAP)估计推导出DCVS中非关键帧重构的优化方程,再基于近端梯度算法推导出优化方程的求解框架,包含多信息流梯度更新聚合方程.基于此,本文设计了多信息流梯度更新及聚合模块(Multi-Information flow Gradient update and Aggregation,MIGA),并构建了深度多信息流梯度更新与聚合网络(Deep Multi-Information flow Gradient update and Aggregation Network,DMIGAN)用于DCVS非关键帧重构.MIGA利用测量值与多参考帧对当前非关键帧进行并行梯度更新,再做信息交互融合,从而充分结合多种信息流更新重构帧.本文级联MIGA与去噪子网络用于模拟近端梯度算法的单次迭代,作为基础模块(phase),并通过级联多个phase构造深度重构网络DMIGAN,实现帧重构的深度优化过程.实验表明,DMIGAN与具代表性的传统迭代优化算法结构相似的帧间组稀疏表示重构算法(Structural SIMilarity based Inter-Frame Group Sparse Representation,SSIM-Inter F-GSR)相比,在低采样率与高采样率下性能分别提升了8.8 dB和7.36 dB;和具有代表性的深度学习重构算法VCSNet-2相比,在低采样率和高采样率下性能分别提升了7.09 dB和8.78 dB.展开更多
本文论证了超分辨率图像复原计算中的两个性质,并基于此在MAP(Maximum A Posteriori)框架下提出了一种新的纹理自适应算法.算法首先根据低分辨率图像和高分辨率图像近似计算的可类比性质计算初始图像,使初始图像的质量更高,并根据超分...本文论证了超分辨率图像复原计算中的两个性质,并基于此在MAP(Maximum A Posteriori)框架下提出了一种新的纹理自适应算法.算法首先根据低分辨率图像和高分辨率图像近似计算的可类比性质计算初始图像,使初始图像的质量更高,并根据超分辨率复原图像阶跃边缘的陡坡性质,将三边滤波正则化应用于迭代运算中,更好地保护了图像的陡坡和屋顶边缘.算法可根据图像的纹理自动计算初始图像融合参数以及正则化函数中的梯度阈值等参数,解决了以往超分辨率图像复原算法参数调整复杂的问题.实验结果表明,本文算法在没有人工参与的情况下,重建图像的客观评价和主观质量均有明显提高.展开更多
提出一种预估计混叠度的PEMAP(pre-estimated MAP (maximum a posteriori))算法,用于卫星图像的地面超分辨率处理.它通过频域分析确定卫星图像的混叠度,将其作为先验信息在空域控制MAP估计的循环迭代,联合估计帧间位移和高分辨率图像....提出一种预估计混叠度的PEMAP(pre-estimated MAP (maximum a posteriori))算法,用于卫星图像的地面超分辨率处理.它通过频域分析确定卫星图像的混叠度,将其作为先验信息在空域控制MAP估计的循环迭代,联合估计帧间位移和高分辨率图像.该算法克服了最大后验概率MAP算法的盲目性和不稳定性,使其适应性更好.实际的卫星图像处理显示了较好的处理效果.展开更多
文摘贝叶斯网络(bayesian network,BN)小数据集条件下,定性最大后验概率(qualitative maximum a posteriori,QMAP)估计往往会违反给定的专家约束,这就导致QMAP估计偏离真实值。为了克服该算法的缺陷,提出了一种改进的QMAP算法。首先,学习出QMAP估计,再结合保序回归方法对违反不等式约束的参数进行校正;然后使用一种微调策略对校正后的参数做进一步调整,使所得参数能够满足专家约束;最后,与最大似然估计(maximum likelihood estimation,MLE)和QMAP算法对比。仿真实验结果表明:在小数据集条件下,提出的算法满足所有约束条件,KL(Kullback-Leibler)散度始终低于其他2种算法,运行时间高于其他2种算法约0.1 s,影响甚微,且推理结果贴近真实值,偏差维持在±0.05之间。改进的QMAP算法的综合性能优于MLE、QMAP算法,并具有较好的实用性。
基金Supported by the National Natural Science Foundation of China(61405191)
文摘A maximum a posteriori（ MAP） algorithm is proposed to improve the accuracy of super resolution（ SR） reconstruction in traditional methods. The algorithm applies both joints image registration and SR reconstruction in the framework,but separates them in the process of iteratiion. Firstly,we estimate the shifting parameters through two lowresolution（ LR） images and use the parameters to reconstruct initial HR images. Then,we update the shifting parameters using HR images. The aforementioned steps are repeated until the ideal HR images are obtained. The metrics such as PSNR and SSIM are used to fully evaluate the quality of the reconstructed image. Experimental results indicate that the proposed method can enhance image resolution efficiently.
文摘The paper presents an algorithm of automatic target detection in Synthetic Aperture Radar(SAR) images based on Maximum A Posteriori(MAP). The algorithm is divided into three steps. First, it employs Gaussian mixture distribution to approximate and estimate multi-modal histogram of SAR image. Then, based on the principle of MAP, when a priori probability is both unknown and learned respectively, the sample pixels are classified into different classes c = {target,shadow, background}. Last, it compares the results of two different target detections. Simulation results preferably indicate that the presented algorithm is fast and robust, with the learned a priori probability, an approach to target detection is reliable and promising.
文摘Interferogram noise reduction is a very important processing step in Interferometric Synthetic Aperture Radar(InSAR) technique. The most difficulty for this step is to remove the noises and preserve the fringes simultaneously. To solve the dilemma, a new interferogram noise reduction algorithm based on the Maximum A Posteriori(MAP) estimate is introduced in this paper. The algorithm is solved under the Total Generalized Variation(TGV) minimization assumption, which exploits the phase characteristics up to the second order differentiation. The ideal noise-free phase consisting of piecewise smooth areas is involved in this assumption, which is coincident with the natural terrain. In order to overcome the phase wraparound effect, complex plane filter is utilized in this algorithm. The simulation and real data experiments show the algorithm can reduce the noises effectively and meanwhile preserve the interferogram fringes very well.
文摘In this paper we consider three problems in continuous multi-criteria optimization: An application of the Berge Maximum Theorem, properties of Pareto-retract mappings, and the structure of Pareto sets. The key goal of this work is to present the relationship between the three problems mentioned above. First, applying the Maximum Theorem we construct the Pareto-retract mappings from the feasible domain onto the Pareto-optimal solutions set if the feasible domain is compact. Next, using these mappings we analyze the structure of the Pareto sets. Some basic topological properties of the Pareto solutions sets in the general case and in the convex case are also discussed.
文摘现有优秀的基于深度学习的分布式视频压缩感知(Distributed Compressed Video Sensing,DCVS)重构算法利用测量值和参考帧顺序更新非关键帧,获得了较好的重构性能,但由于缺乏较严格的理论指导,无法充分结合这两类信息,限制了非关键帧重构质量的进一步提升.针对该问题,本文首先利用贝叶斯理论及最大后验概率(Maximum A Posteriori,MAP)估计推导出DCVS中非关键帧重构的优化方程,再基于近端梯度算法推导出优化方程的求解框架,包含多信息流梯度更新聚合方程.基于此,本文设计了多信息流梯度更新及聚合模块(Multi-Information flow Gradient update and Aggregation,MIGA),并构建了深度多信息流梯度更新与聚合网络(Deep Multi-Information flow Gradient update and Aggregation Network,DMIGAN)用于DCVS非关键帧重构.MIGA利用测量值与多参考帧对当前非关键帧进行并行梯度更新,再做信息交互融合,从而充分结合多种信息流更新重构帧.本文级联MIGA与去噪子网络用于模拟近端梯度算法的单次迭代,作为基础模块(phase),并通过级联多个phase构造深度重构网络DMIGAN,实现帧重构的深度优化过程.实验表明,DMIGAN与具代表性的传统迭代优化算法结构相似的帧间组稀疏表示重构算法(Structural SIMilarity based Inter-Frame Group Sparse Representation,SSIM-Inter F-GSR)相比,在低采样率与高采样率下性能分别提升了8.8 dB和7.36 dB;和具有代表性的深度学习重构算法VCSNet-2相比,在低采样率和高采样率下性能分别提升了7.09 dB和8.78 dB.
文摘本文论证了超分辨率图像复原计算中的两个性质,并基于此在MAP(Maximum A Posteriori)框架下提出了一种新的纹理自适应算法.算法首先根据低分辨率图像和高分辨率图像近似计算的可类比性质计算初始图像,使初始图像的质量更高,并根据超分辨率复原图像阶跃边缘的陡坡性质,将三边滤波正则化应用于迭代运算中,更好地保护了图像的陡坡和屋顶边缘.算法可根据图像的纹理自动计算初始图像融合参数以及正则化函数中的梯度阈值等参数,解决了以往超分辨率图像复原算法参数调整复杂的问题.实验结果表明,本文算法在没有人工参与的情况下,重建图像的客观评价和主观质量均有明显提高.
文摘提出一种预估计混叠度的PEMAP(pre-estimated MAP (maximum a posteriori))算法,用于卫星图像的地面超分辨率处理.它通过频域分析确定卫星图像的混叠度,将其作为先验信息在空域控制MAP估计的循环迭代,联合估计帧间位移和高分辨率图像.该算法克服了最大后验概率MAP算法的盲目性和不稳定性,使其适应性更好.实际的卫星图像处理显示了较好的处理效果.
