The invention discloses an automatic crack detection method based on a seismic coherence cube slice and relates to the technical field of petroleum exploration image processing. The method comprises the following steps of (S1) median filtering and denoising, (S2) normalization processing, (S3) bilateral filtering and Laplace filtering processing, (S4) the removal of isolated noise with connected component marking, (S5) morphological operation processing, (S6) the repair of a 2D seismic coherence cube slice with a flood water filling algorithm, (S7) the extraction of a framework by using an FPAalgorithm, (S8) the enhancement of a crack continuity with a MyPadding function, (S9) the marking of the connected component, (S10) processing by using a least squares straight line fitting method, and (S11) classification display including the step of classifying cracks according to the range of different azimuth angles to obtain a final crack automatic detection result of 2D seismic coherence cube slice data. According to the method, the seismic coherence cube slice data can be directly used to extract crack information, the calculation of the azimuth of the cracks is achieved, the crack interpretation efficiency of the seismic slice is greatly improved, and a lot of manpower and material resource costs are saved.