The invention relates to an omnidirectional search mechanism variogram based optimal remote sensing scale selection method, which comprises the following steps of: firstly, selecting a research area by using an obtained remote sensing image; then, calculating corresponding variogram values of the remote sensing image at different space intervals in each direction by using an omnidirectional search algorithm; on this basis, establishing a curvilinear relation between the space intervals and the variogram values of the space intervals, and then solving a corresponding local optimal space scale of the remote sensing image in each direction based on the spherical model regression fitting in a variogram model; and finally, taking the local optimal space scale corresponding to the direction with the maximum degree of fitting as the optimal space scale of the remote sensing image. By using the method disclosed by the invention, the problem of 'false' local optimal scales caused by that when a traditional variogram algorithm is used, in the process of selecting the optimal space scale of a remote sensing image, only an isotropic factor is taken into consideration is solved, thereby improving the accuracy of optimal scale selection.