摘要
互信息率(MIR)作为随机序列统计独立性的一种度量,同时包含了序列的幅度信息与相位信息,特别适合作为盲反褶积的统计准则.但基于MIR的盲反褶积方法对噪声非常敏感,并且该算法的运算量很大.本文研究了随机噪声对反褶积的影响,通过在目标函数中加入噪声约束项,抑制了反褶积过程中高频噪声的干扰,加强了算法的抗噪性能.针对算法计算量大的缺点,文中提出一种新的初值选择方法,计算效率比原方法提高了8～10倍.合成记录与实际地震资料的处理结果表明,该方法可适应于非最小相位子波,能够有效拓展地震记录的频带,提高地震资料的分辨率.
As a statistical metric, mutual information rate （MIR） includes both amplitude and phase information of random sequence, so it is a good criterion for blind deconvolution. However, the noise suppressing capability of MIR based deconvolution is very weak. In addition, this method has a considerable amount of calculation. Since the deconvolution based on MIR is sensitive to noise, we add a regularization term, called noise-correction term, in the objective function of deconvolution, to weaken the high frequency interference. And in order to reduce calculation, a new method of initial value selection is proposed in this paper, it can decrease the iteration of optimization algorithm about 10 times by contrast with original method. Tests on synthetic signal demonstrate that the improved deconvolution method can be used in non-minimum phase system. And result of real data processing shows that this method can effectively expand the frequency band of seismic records, and enhance the resolution of seismic data.
出处
《地球物理学进展》
CSCD
北大核心
2011年第2期540-548,共9页
Progress in Geophysics
基金
国家自然科学基金重点项目(40730424)
国家863计划项目(2006AA09A102)
国家科技重大专项项目(2008ZX050442-6-1)资助
关键词
互信息率
盲反褶积
抗噪性能
计算效率
mutual information rate, blind deconvolution, anti-noise capability, computational efficiency
