The invention discloses a spherical video image correction method, which belongs to the technical field of video processing, and is characterized by comprising the following steps of: a, determining adisplay field angle parameter of each channel; b, generating a multi-channel rectangular graph, and synchronously outputting the multi-channel rectangular graph to a rear-end graph corrector for geometric correction of the graph from a plane to a spherical surface; c, converting each channel image into a curved surface image by the graph corrector; d, storing the graphic correction parameters inFLASH; and e, projecting the pre-deformed image onto a hemispherical projection screen by the projector model for presenting a real target image. Geometric transformation from a plane image to a spherical image can be achieved by loading the parameters through hardware electrification, and image acquisition, grid correction, feature point acquisition and calculation of a limited display area are not needed; effective pixels of a video image are fully utilized, real-time correction processing of the video image is achieved, reusability of parameters between channels is enhanced, and secondary correction of later maintenance is avoided.