中文摘要：

本文给出了地震记录变子波模型的一种近似数学表达式．基于该表达式研究了反射系数序列不满足白噪假设和子波在地下传播时发生变化这两种情况下地震道谱的组成及结构，讨论了谱白化及反褶积方法在这两种情况下效果不佳的原因．然后基于变子波模型，提出了一种新的提高地震记录分辨率的方法：第一步，用自适应于地震记录的Gabor分子窗把地震记录恰当地划分成若干片断，每段内信号近似平稳，然后将地震记录变换到时间一频率域；第二步，在变换域对每个分子窗内信号的振幅谱进行处理以拓宽频带；最后把处理后的时间一频率域函数反变换回时间域得到提高分辨率后的结果．本文提出的方法具有能较好地适用于反射系数不满足白噪假设的情况及提高分辨率后的地震记录能较好地保持原地震记录的相对能量关系等优点，模型和实际资料算例结果均表明，本文方法在拓宽地震资料频带及保持地震记录局部能量相对关系方面均明显优于谱白化方法．

英文摘要：

We present a mathematic expression of the changing wavelet model of the seismogram （CWMS）. Based on this expression, we study the content and structure of spectra of seismic traces produced in two cases, i.e. the reflectivity does not meet the white assumption and seismic wavelets change during travel in subsurface. And we analyze why the spectral-whitening and deconvolution methods cannot yield good results in these two cases. Then based on CWMS, we propose a new method to enhance the resolution of seismic traces. Firstly, we divide the seismic signal into proper segments by Gabor molecule-windows adaptive to the signal, where each segment is regarded stationary, and then transform it into the Gabor domain. Secondly, we deal with the amplitude spectrum for each segment to expand the frequency band of the spectrum in the Gabor domain. Finally, taking inverse Gabor transform on modified time-frequency coefficients, we get the high-resolution signal. Compared with the spectral-whitening and deconvolution methods, our method have the following advantages： （1） it can work well even when the reflectivity doesn＇t satisfy white presumption; （2） it can preserve local energy relationship with original seismic trace which may be important for seismic data interpretation. Tests on synthetic and real data show that our method is much better than the spectral-whitening method in expanding frequency bandwidth and preserving local energy relationship of seismograms.