中文摘要：

稀疏化方法由于能够实现地震信号的高精度分解，已经成为重要的地震信号处理技术．目前地震信号稀疏分解常采用的方法是匹配追踪算法（MatchingPursuit，MP），但所得结果不够稀疏．针对此局限，提出了一种基于重复加权提升搜索算法（RepeatedWeightedBoostingSearch，RWBS）的快速分解方法．首先，根据地震信号的频谱图缩小频率搜索范围；然后，将搜索算法RWBS与正交匹配追踪（OrthogonaIMatchingPursuit，OMP）方法相结合，就得到一种快速的稀疏分解方法．将本文的方法应用到人工合成和实际的地震数据处理中，并与MP和OMP追踪算法作比较，说明采用本文方法进行地震信号分解在稀疏度和分解速度方面都有提高．仿真实验结果表明，与MP和OMP分解算法相比，在满足相同的分解精度条件下，RWBS算法不仅大大提高了分解的稀疏度，而且提高分解速度．与OMP算法相比较，基于RWBS的新方法分解所需的时间减少了约87％；与MP算法相比较，新方法分解所需的时间减少约50％．

英文摘要：

Sparse representation has become an important seismic processing technology. Matching pursuit （ MP ） algorithm is often used in sparse decomposition of seismic signal, but the resulting method is not sparse. Another common algorithm , orthogonal matching pursuit （OMP） algorithm, is much sparser than MP, but the complexity of OMP algorithm is too high. In order to overcome those problems, we propose a rapid method of decomposition, which is combined repeated weighted boosting search （RWBS） with OMP. The search space is significantly reduced by investigating the frequency scope of given seismic records. Application of the proposed method to synthetic data and real data of seismic, and comparison of the results with other used decomposition approaches of MP and OMP, illustrate the ability of our approach to provide seismic data decomposition results highly localized in both sparsity and running time. Numerical simulations show that, algorithm based on RWBS performed much better than the traditional ones in terms of running time, sparsity and accuracy. Competed with OMP algorithm, the running time of new algorithm based on RWBS is reduced by about 87% compared with MP algorithm, it is declined by about 50%.