According to noise jamming existing in ultraviolet imaging process, typical UV imaging system are researched for acquiring high quality UV images and inspecting targets accurately. The reasons that noise jamming exists in ultraviolet imaging process are analyzed. The scene models of ultraviolet targets are established. Based on this, the restoration methods of ultraviolet images are researched. Single image is performed threshold processing so as to cut out background noise. Then multi-frame continuous images are performed average processing and time-recursive filtering. Better image output is got after threshold segmentation process. Based on output results of images, the random noise from ultraviolet images can be removed effectively by multi-frame image average processing and time-recursive filtering. Especially the method of time-recursive filtering has better effect on image restoration.