Abstract: The statistical feature of the remote sensing image plays an important role for image recognition, image registration and image classification. The traditional robustness analysis of statistical feature is generally based on that the remote sensing images have same sensor parameters. However, this is not true in many practical applications, where the sensors may have different sensor parameters, e.g., wavelength and incidence angle etc. Due to various types of sensor parameters, the remote sensing image will have different appearance. This will further affect the statistical features. Therefore, in order to achieve more robust image processing and analysis, it's necessary to investigate the correlations between the statistical features used in the state of the art and the sensor parameters. In this paper, we research on the robustness of six kinds of statistical features against the wavelength (L-band, C band and X band) of the synthetic aperture radar (SAR). First the measurement of robustness of statistical features is defined, and then the comparison and analysis are presented in our experiment. In addition, our experiment also shows that statistical features of different terrains have different change while wavelength changes. The results of this paper also play a very important role in the SAR image based interpretation, image matching guidance and scene adaptability analysis.