针对基于稀疏不变性假设的单帧超分辨率(SR)算法的局限性,提出一种利用相似最近邻(ANN)统计预测模型的单帧SR算法。首先,利用相似最近邻思想,通过波尔茨曼机捕捉HR字典与LR字典对稀疏模式之间的依赖关系,建立统计预测模型;然后,根据LR块...针对基于稀疏不变性假设的单帧超分辨率(SR)算法的局限性,提出一种利用相似最近邻(ANN)统计预测模型的单帧SR算法。首先,利用相似最近邻思想,通过波尔茨曼机捕捉HR字典与LR字典对稀疏模式之间的依赖关系,建立统计预测模型;然后,根据LR块与HR块相关的最小均方误差(MMSE)计算网络参数,获得它们的依赖关系;最后,利用多层前向神经网络提取字典元素内积,通过计算重叠局部块预测值的均值来重建图像。利用峰值信噪比PSNR和结构相似性度量SSIM评估实验结果,实验结果表明,提出的算法在视觉效果和数值标准方面大多优于其他算法,在选择合适参数情况下,峰值信噪比至少提高0.2 d B。展开更多
A new multi-sensor data fusion algorithm based on EMD-MMSE was proposed.Empirical mode decomposition(EMD)is used to extract the noise of every time series for estimating the variance of the noise.Then minimum mean squ...A new multi-sensor data fusion algorithm based on EMD-MMSE was proposed.Empirical mode decomposition(EMD)is used to extract the noise of every time series for estimating the variance of the noise.Then minimum mean square error(MMSE)estimator is used to calculate the weights of the corresponding series.Finally,the fused signal is the weighted addition of all these series.The experiments in lab testified the efficiency of this method.In addition,the comparison in fusion time and fusion results with existing fusion method based on wavelet and average technique shows the advantage of this method greatly.展开更多
文摘针对基于稀疏不变性假设的单帧超分辨率(SR)算法的局限性,提出一种利用相似最近邻(ANN)统计预测模型的单帧SR算法。首先,利用相似最近邻思想,通过波尔茨曼机捕捉HR字典与LR字典对稀疏模式之间的依赖关系,建立统计预测模型;然后,根据LR块与HR块相关的最小均方误差(MMSE)计算网络参数,获得它们的依赖关系;最后,利用多层前向神经网络提取字典元素内积,通过计算重叠局部块预测值的均值来重建图像。利用峰值信噪比PSNR和结构相似性度量SSIM评估实验结果,实验结果表明,提出的算法在视觉效果和数值标准方面大多优于其他算法,在选择合适参数情况下,峰值信噪比至少提高0.2 d B。
基金The National High Technology Research and Development Program of China(863Program)(No.2001AA602021)
文摘A new multi-sensor data fusion algorithm based on EMD-MMSE was proposed.Empirical mode decomposition(EMD)is used to extract the noise of every time series for estimating the variance of the noise.Then minimum mean square error(MMSE)estimator is used to calculate the weights of the corresponding series.Finally,the fused signal is the weighted addition of all these series.The experiments in lab testified the efficiency of this method.In addition,the comparison in fusion time and fusion results with existing fusion method based on wavelet and average technique shows the advantage of this method greatly.
