Abstract: To solve the problem that the traditional methods are difficult to detect the seismic fault in case of the effective seismic horizons in the section are mixed with noises, a fault recognition method based on connected component dispersion was proposed. First, the connected component in the binary seismic image was labeled out based on the eight neighbor connectivity and the digital dispersion and other characteristics of each connected component were extracted. Then, the noises were eliminated based on the horizontal length and dispersion difference between the noise and the connected component corresponding to the seismic horizons. Furthermore, the seismic faults were identified and labeled out based on the horizontal endpoints of the connected component corresponding to the seismic horizons. Finally, the algorithm was compared with the traditional ones through experiments, and the results validated the effectiveness and the practicability of the proposed algorithm.