In this paper, we define the harmonic oscillator with random damping in non-Markovian thermal bath. This model represents new version of the random oscillators. In this side, we derive the overdamped harmonic oscillat...In this paper, we define the harmonic oscillator with random damping in non-Markovian thermal bath. This model represents new version of the random oscillators. In this side, we derive the overdamped harmonic oscillator with multiplicative colored noise and translate it into the additive colored noise by changing the variables. The overdamped harmonic oscillator is stochastic differential equation driving by colored noise. We derive the change in the total entropy production (CTEP) of the model and calculate the mean and variance. We show the fluctuation theorem (FT) which is invalid at any order in the time correlation. The problem of the deriving of the CTEP is studied in two different examples of the harmonic potential. Finally, we give the conclusion and plan for future works.展开更多
针对彩色图像边缘提取算法产生伪边缘和间断边缘的不足,提出自适应向量总变分与色彩差异的彩图边缘提取方案。根据图像梯度值与幅值,定义一种自适应向量总变分(vector total variation,VTV)去噪模型,降低噪声对边缘的影响;将图像变换为C...针对彩色图像边缘提取算法产生伪边缘和间断边缘的不足,提出自适应向量总变分与色彩差异的彩图边缘提取方案。根据图像梯度值与幅值,定义一种自适应向量总变分(vector total variation,VTV)去噪模型,降低噪声对边缘的影响;将图像变换为CIELAB空间,计算其色差和方向;对于不同色差与方向,采用不同的Sobel算子,借助非最大抑制方法(non-maximum suppression,NMS)优化边缘,搜索更多边缘点;利用自适应双阈值(double threshold,DT)提取彩色图像边缘。实验结果表明,与当前图像边缘提取技术相比,所提算法具有更优的检测结果,以及更高的边缘评价因子F与更低的Baddeley误差。展开更多
现有的大多数单图像超分辨率方法仅用于提高单个通道的分辨率。在处理彩色图像时,由于忽略了通道间的相关性,重建的高分辨率图像容易产生失真。针对这些问题,提出了一种综合考虑通道间相关性及非局部自相似性的彩色图像超分辨算法。首先...现有的大多数单图像超分辨率方法仅用于提高单个通道的分辨率。在处理彩色图像时,由于忽略了通道间的相关性,重建的高分辨率图像容易产生失真。针对这些问题,提出了一种综合考虑通道间相关性及非局部自相似性的彩色图像超分辨算法。首先,为了充分利用彩色图像的通道间相关性,分别计算通道间残差信号和三通道平均信号的总变分范数;其次,为了进一步提升超分辨率的结果,基于图像内的非局部自相似性更新重建图像;最后,为了求解所建立的优化问题,提出了基于split-Bregman方法的快速迭代算法。将所提算法与一些主流算法进行了比较,在3倍上采样条件下,所提算法在Set5和Set14数据集上平均可获得的峰值信噪比(Peak Signal to Noise Ratio,PSNR)增益分别为0.5 dB及0.36 dB。实验结果证明了联合应用通道间相关性及非局部自相似性能有效提升彩色图像的超分辨重建质量。展开更多
文摘In this paper, we define the harmonic oscillator with random damping in non-Markovian thermal bath. This model represents new version of the random oscillators. In this side, we derive the overdamped harmonic oscillator with multiplicative colored noise and translate it into the additive colored noise by changing the variables. The overdamped harmonic oscillator is stochastic differential equation driving by colored noise. We derive the change in the total entropy production (CTEP) of the model and calculate the mean and variance. We show the fluctuation theorem (FT) which is invalid at any order in the time correlation. The problem of the deriving of the CTEP is studied in two different examples of the harmonic potential. Finally, we give the conclusion and plan for future works.
文摘针对彩色图像边缘提取算法产生伪边缘和间断边缘的不足,提出自适应向量总变分与色彩差异的彩图边缘提取方案。根据图像梯度值与幅值,定义一种自适应向量总变分(vector total variation,VTV)去噪模型,降低噪声对边缘的影响;将图像变换为CIELAB空间,计算其色差和方向;对于不同色差与方向,采用不同的Sobel算子,借助非最大抑制方法(non-maximum suppression,NMS)优化边缘,搜索更多边缘点;利用自适应双阈值(double threshold,DT)提取彩色图像边缘。实验结果表明,与当前图像边缘提取技术相比,所提算法具有更优的检测结果,以及更高的边缘评价因子F与更低的Baddeley误差。
文摘现有的大多数单图像超分辨率方法仅用于提高单个通道的分辨率。在处理彩色图像时,由于忽略了通道间的相关性,重建的高分辨率图像容易产生失真。针对这些问题,提出了一种综合考虑通道间相关性及非局部自相似性的彩色图像超分辨算法。首先,为了充分利用彩色图像的通道间相关性,分别计算通道间残差信号和三通道平均信号的总变分范数;其次,为了进一步提升超分辨率的结果,基于图像内的非局部自相似性更新重建图像;最后,为了求解所建立的优化问题,提出了基于split-Bregman方法的快速迭代算法。将所提算法与一些主流算法进行了比较,在3倍上采样条件下,所提算法在Set5和Set14数据集上平均可获得的峰值信噪比(Peak Signal to Noise Ratio,PSNR)增益分别为0.5 dB及0.36 dB。实验结果证明了联合应用通道间相关性及非局部自相似性能有效提升彩色图像的超分辨重建质量。
