The most time-consuming part of the conventional Shooting and Bouncing Ray (SBR) method is the ray tracing process, which increases the computation time because of the large number of ray tubes used to model the incident wave and the large number of triangles used to model the surface mesh. In this study, an accelerated SBR method is presented to reduce the computational burden. Acceleration is done both by reducing the amount of ray-triangle intersection tests with the help of octree decomposition of 3D mesh structure, and by parallelizing the ray tracing algorithm on GPU (graphics processing unit) in MATLAB. The proposed method is used to obtain the Inverse Synthetic Aperture Radar (ISAR) image of electrically-large complex objects. The numerical results are compared with those of a commercial software.