ABSTRACT
Relative radiometric normalization helps to minimize the effect of solar illumination conditions and atmospheric attenuation on images based on a reference image. In this chapter six statistical relative radiometric normalization (SRRN) techniques were applied and compared to normalize the bi-temporal Sentinel-2A multi-spectral data. The target image of the year 2019 was normalized in such a way that it resembled the similar atmospheric and sensor conditions available on the reference image of the year 2016. For this purpose, similar seasonal images were used. The applied techniques were compared and based on the error statistic; it was found that the Iteratively Re-weighted Multivariate Alteration Detection (IR-MAD) method performed better. The IR-MAD technique is based on the Iterative statistical Canonical Component Analysis (ICCA). It is advantageous as it identifies the true set of time-invariant pixels between the images using this ICCA. The effect of normalization is more pronounced on high spatial resolution bands of the Sentinel 2A image. Other than IR-MAD, the simple haze correction technique yielded acceptable results. The IR-MAD normalized image stretches the values of the well-known Normalized Difference Vegetation Index and recently developed Combined Mangrove Recognition Index. The regression analysis and one-way ANOVA test supports this finding. The result of this study may help to estimate and monitor the types and health of mangrove vegetation of the Indian Sundarbans in a better possible way.
AU: Please provide the missing keywords for this chapter.
