In order to solve the restoration problem of images with radial blur and random vibration blur, which is caused by EO payload imaging in high speed motion, a physical model of hybrid blurred image is established by using the sub image blur kernel function extension estimation step by step, to improve the RL TV recursive model and complete the restoration of hybrid blurred image. The results show that the hybrid blur kernel function is a superimposition of the vibration vector and radial vector in time of exposure, and the simulation effect conforms with the actual measured effect. In the hybrid blurred image, the radial blurred features gradually attenuate from the edge to the center, and the central area of image can approximately estimate the vibration blurred kernel function. By using a center to edge sub image hierarchical recursive method, the hybrid blurred image can be effectively restored, with much better effect than other restoration methods, such as sequence restoration.