During the process of developing directional polarimetric camera, its stability is an important performance parameter, which represents the normal working state of the imaging remote sensing instruments when it is in orbit. Therefore, it is extremely important to accurately measure the stability parameters of the imaging remote sensing instruments, which can provide instructive information for developing imaging remote sensing instruments. At present, the traditional halogen tungsten lamp integrating sphere light source is usually adopted as a radiation source. At the same time, a wide-spectrum trap detector is used to synchronously monitor the light energy stability of the light source, and the obtained monitoring data is used for stability parameter processing correction. In order to monitor the stability of the integrating sphere radiation in different wavelength bands better and solve the problems of spectral bandwidth mismatch as well as data acquisition time alignment difficulty, which cannot ensure measurement accuracy, the method is proposed in this paper. It works by using data of directional polarimetric camera data instead of trap detector data, 12 channels of directional polarimetric camera data wavelet decomposition, respectively, to extract the variation of each band energy. Then normalize the fluctuation of each band energy. It can be inferred that the change of the digital number caused by the fluctuation of the traditional halogen tungsten lamp integrating sphere light source. Therefore, through measurement data correction, the influence of light source instability can be deducted and effectively improve the stability parameter measuring accuracy. After deducting the light source fluctuation according to the above method, the instability parameter of the directional polarimetric camera has reduced from 0.153% to 0.031%. The corrected value shows the actual instability of the directional polarimetric camera more realistically. The detected light source energy variation trend of the same band with three detection directions is highly consistent, and the signal-to-noise ratio of the multi-frame image method is improved more significantly, which strongly proves the effectiveness of the proposed method. Compared with the signal-to-noise ratio before being corrected, the corrected signal-to-noise ratio has increased by 2.52% on average by the wavelet decomposition de-illumination method, while only increased by 0.72% on average by the trap detector de-illumination. In the 865 nm and 910 nm bands, the wavelet decomposition de-illumination method improves the signal-to-noise ratio by 4.34% and 12.9%. It is worth noting that the signal-to-noise ratio increase in the correction by wavelet decomposition de-illumination method is much larger than that of the other method. It also shows that there are limitations in correcting fluctuation in the 865 nm and 910 nm bands with broad-spectrum trap detector monitoring data. In the 490 nm band, the wavelet decomposition de-illumination method increased by 0.54%, but the other correction method only increased by 0.09%, mainly because the light source energy in the 490 nm band is very low, so the digital number of directional polarimetric camera acquired is the lowest. Therefore, although the light source has the largest change in the 490 nm band, the improvement of the signal-to-noise ratio is not as significant as that in the 910 nm band. The reason for the small signal-to-noise ratio improvement by the two correction methods in 670 nm band is that the variation of the light source in this band is less than 0.205%. Therefore, it is reasonable that the improvement effect is not obvious. The proposed method can be used to test the stability parameters of various imaging remote sensing instruments.