A fast computation algorithm for aerial images is presented based on the sysmmetric properties of illumination light and matrix multipication. The new discretization approach divides the light source into many concentric circles based on the sysmmetric properties, and the resulting sampled source points can be put along one line outgoing from the origin point, which can reduce the sampling number due to its simplified two dimension traditional orthogonal discretization method to one dimension concentric circles method. In addition, a matrix multiplication technique is adopted to speed up the computation. The simulation results show that under the same sampling number, the circle discretization approach has a higher accuracy than the traditional orthogonal discretization method, and under the same accuracy, the sampling number of concentric circles method is 1/N to the traditional orthogonal discretization method (N×N).