Analyzing the Number of Looks from Stochastic Distance in Polarimetric SAR Imagery

The concept of multi-looking has a considerable significance in polarimetric synthetic aperture radar imagery, whether it is to remove distortions or obtain square pixels from rectangular ones. This paper aims at estimating the optimal number of looks, given a coherency matrix. Our methodology involves finding the Hellinger distance between a rotated coherency matrix and the original coherency matrix for a set of polarimetric orientation angles. Then, as we obtain a plot between δ
H
(difference between optimized T22 and T33 at 15
⁰
or -15
⁰
) and the number of looks, we perform lognormal curve fitting on it. By the definition of a lognormal distribution, the mean of the distribution is the optimized number of looks, and the value of the distribution is the δ
H
value.