【摘要】 对经验数据分解算法进行了研究,从滤波器完全重构出发,研究了经验数据分解滤波器与两通道滤波器组对应滤波器参数的关系,得出分析和综合滤波器都为FIR滤波器应满足的条件。可根据不同的应用对分析高通滤波器进行灵活地设计,并保证重建滤波器是FIR滤波器。利用离散线性定常系统的最优滤波和预测技术对分析高通滤波器的系数进行了预测,使用超光谱图像数据进行测试分析,确定了最佳预测滤波器的系数。结合JPEG2000的核心算法EBCOT对超光谱图像数据进行处理,并比较了该算法的输出码率同JPEG2000中无损压缩输出码率。结果表明经验数据分解算法产生的图像输出码率比JPEG2000中无损压缩输出码率平均小0.217bpp,是一种对非平稳数据有效的分析方法。更多还原

【Abstract】 The algorithm of Empirical Data Decomposition(EDD) is been studied.In terms of reconstructing requirement,the relation between the parameters of EDD filters and those of bi-orthogonal filter banks is analyzed and the condition that the analysis filter and the synthesis filter are FIR filters is discussed.The analysis high-pass filter can be feasibly designed according to different applications and at the same time the reconstructing filter is surely FIR filter.The coefficients of the analysis high-pass filter are predicted by use of optimal filtering and predicting technology of discrete linear system.The optimal predicting filter coefficients are decided through testing on spectrum image data.The spectrum image data is processed by the algorithm proposed in the paper and the kernel algorithm of JPEG2000,that is Embedded Block Coding with Optimized Truncation(EBCOT).The comparison between the output code rate of the EDD algorithm and that JPEG2000 loss-less compression shows the output code rate of lossless compression declines 0.217bpp on the average with EDD algorithm,and proved EDD is an effective analysis method for non-stationary image data.更多还原