中文摘要：

针对单一型数学变换或字典不能有效刻画地震信号的形态特征多样性这一问题,在形态分量分析（MCA）框架下,提出了一种基于离散余弦变换（DCT）与曲波字典组合的地震信号重建方法。该方法首先建立MCA框架下的地震信号重建模型,依托模型将信号分解成局部奇异形态分量以及平滑与线状形态分量。然后采用DCT字典表示局部奇异分量,采用曲波字典表示平滑与线状分量。再以迭代求解方式逐一重建各分量,最后将重建后的分量合并。人工合成地震信号和二维叠前及叠后实际地震信号重建实验结果表明,该方法能很好完成信号重建,重建精度不仅要高于非抽样小波变换（UDWT）与曲波字典组合、曲波与曲波字典组合、余弦调制滤波器组与曲波字典组合,而且更高于DCT,UDWT,或曲波等单一型字典。

英文摘要：

Aiming at the problem that the mathematic transforms and dictionaries can not effectively depict the morphological features diversity of seismic signals, we propose a seismic signal reconstruction method under the morphological component a- nalysis （MCA） framework combined with discrete cosine transform （DCT） and curvelet dictionary. Firstly the seismic signal reconstruction model is built under the MCA framework. Then the signal is decomposed into local singular and smooth linear component based on the model. Following that, the local singular component is represented by DCT dictionary, and the smooth linear component is represented by curvelet dictionary. We combine two kinds of components together after iterative reconstruction. The experiments on synthetic and real seismic signals illustrates that the proposed method can be used to re- construct signals very well. The reconstruction precision of the method is not only higher than some dictionary combinations such as UDWT ＆ curvelet dictionary combination, curvelet ＆ curvelet dictionary combination,CMBF ＆ curvelet dictionary combination, but also higher than some single dictionaries such as DCT, UDWT or curvelet etc.