Normal moveout correction is crucial in seismic data processing, but it generates a wavelet-stretching effect, especially on the larger offset or incident-angle seismic data. Wavelet stretching reduces the dominant frequency of seismic data. The greater the incident angle or offset, the lower dominant the frequency becomes. This is an unfavorable effect to amplitude variation with offset analysis. Therefore, we have introduced a wavelet stretching correction method based on the multiwavelet decomposition (MWD) algorithm. First, it decomposes the near-offset pilot trace and all the far-offset seismic traces in the same gather into a series of wavelets via the MWD algorithm. Then, the dominant frequencies of wavelets in the far-offset seismic traces are replaced by those corresponding wavelets in the pilot trace. Finally, the wavelets after the stretching correction are used to reconstruct the seismic trace. The model and field-data processing results show that this method can not only effectively reduce the wavelet stretching effect but it can also maintain the amplitude of each wavelet as invariant during the stretching correction procedure. Because only the frequencies of the decomposed wavelets are used, and no inverse wavelet operators is introduced, the wavelet stretching correction method does not distort the amplitude information.