The invention relates to a high-dimensional wavelet transform-based method for discontinuity detection of seismic data. The method comprises the following steps: (1), collecting seismic data and carrying out pretreatment to obtain a three-dimensional seismic data volume; (2), carrying out Hilbert transform on each channel of the three-dimensional seismic data volume so as to obtain a three-dimensional phase cosine data volume; (3), carrying out partitioning on the three-dimensional phase cosine data volume and selecting searching scopes of scale, inclination and azimuth according to discontinuous scales and geographical objects; (4), carrying out three-dimensional continuous wavelet transform on each subblock of the three-dimensional phase cosine data volume; (5), giving discontinuity tolerances of the seismic data according to a three-dimensional continuous wavelet transform coefficient modulus of each the subblock of the three-dimensional phase cosine data volume so as to obtain a partitioning discontinuous data volume; (6), carrying out splicing on the generated partitioning discontinuous data volume to obtain a discontinuous data volume of the whole seismic data. According to the method provided in the invention, computation quantity is small and multiscale detection can be carried out on discontinuity of seismic data; therefore, the method can widely applied to the detection process of the seismic data.