摘要
We propose a non-dyadic Curvelet transform based least-squares matching method.Firstly,in order to separate the signal and noise better,the input data is decomposed using non-dyadic Curvelet transform according to their spectral and directional characteristics.Secondly,the least-squares matching method is applied on non-dyadic Curvelet coefficients which contain noises most.Our approach improves the stability and accuracy of conventional time-spatial domain least-squares matching method.Results of field records show that the proposed approach performs well not only in suppressing energies of surface waves,but also in protecting effective reflection events,especially in case of overlapping events.The proposed approach can also be used to adaptively subtract predictable interferences such as free surface-related multiples and internal multiples.