The video satellite uses the agile attitude performance to stare imaging for ground achieving the dynamic monitoring of the target. In order to eliminate the non-uniformity of sequence images acquired by the video satellite sensor, which was caused by the response difference of the detection element. To obtain the clear image and the accurate radiation information, relative radiometric calibration was needed for video satellite sensors. Because of the different imaging mechanism between video satellite and push-broom satellite, on-orbit radiometric calibration methods based on the statistics were not suitable for video satellite. The uniform scene on-orbit relative radiometric calibration method to video satellite was proposed, through staring imaging for three typical uniform scenes including cloud, ocean and desert respectively to calculate relative radiometric calibration coefficients. Among them, the processing of desert scene had achieved better relative radiometric calibration coefficients, which was used for correcting single frame image of Jilin-1 video satellite. The method could effectively repair the image defects, and the non-uniformity of the image was reduced from 3.7% to 1.2%.