Since SPOT1 launch in February 1986 and until SPOT5 launch in May 2002, the methods and means to insure the best quality of the images delivered to SPOT IMAGE customers have been continuously improved and updated. The quality of the corrected images is quantified through several figures of merit, including, for radiometric quality, the Signal-to-Noise Ratio (SNR).
Radiometric noise is due to two separate phenomena:
-column-wise noise: it represents on-board image chain performances.
-line-wise noise: normalization defects (radiometric model deviations) may lead to visible “columns” on a uniform landscape.
For each image, these two noises are combined in an “image noise” that quantifies the variations of the digital numbers on a uniform landscape.
Different techniques can be used to assess these different noises in-flight. For SPOT1 to SPOT4, the on-board lamp is used for both normalization and SNR assessment. For SPOT5, without lamp unit, we use images acquired over the quasi-uniform landscapes of Antarctic and Greenland for normalization. However, uniformity of these landscapes is not sufficient to accurately measure the SNR. So, a new method experimented during SPOT3 in-flight commissioning phase and operational for SPOT5 is applied. It consists in using two images of the same landscape acquired simultaneously to eliminate the landscape contribution.
This paper focuses on the presentation of this new method and compares its accuracy to the other methods. Finally, a comparison between flight measurements and ground measurements before launch is given.