The cloud camera adopts the threshold method to detect cloud distribution information in the observation target area. In the cloud judgment algorithm, the accuracy of the radiance will affect the final inversion result. An information model between the response value of the pixel and the radiance was established. Improvement measures are proposed for the situation where the output response values of each pixel of the cloud camera are unequal when the incident light is uniform. First, according to the information model, the reasons for the different response values of the detector were analyzed, and the relative radiometric calibration model was established. It is difficult for the light emitted by the large-diameter integrating sphere radiation source to cover all the fields of view simultaneously due to the large field of the cloud detection camera. Therefore, a subfield of view measurement method was designed. The relative radiometric calibration of the cloud camera was performed using the calibration model and the subfield of view measurement method. The experimental results indicate that the nonuniformity of the imaging results of the integrating sphere is approximately 6% before correction. After the relative radiometric calibration correction, the nonuniformity is less than 1.5% under a typical signal-noise-ratio (SNR) and approximately 3.25% under a low SNR. The results of the calibration meet the requirements of the design of the cloud camera.