Circular Synthetic Aperture Radar (CSAR) is a recently developed all-directional high-resolution imaging mode, which is efficient in dealing with the target recognition, area monitoring, and three-dimensional reconstruction because of the acquisition of 360° data in a single pass. To obtain the entire 360° image using information from all subapertures, the following methods are used: (1) the coherent addition method, (2) the incoherent addition method, and (3) the maximum-intensity methods. In this study, different statistical models of speckle in CSAR images are proposed and the speckle reduction of each model is discussed. Experiments show that the incoherent addition and maximum-intensity methods reduce speckle, whereas the coherent addition method does not.