Equivalent number of looks - quad-pol data

Hello all,

I have consulted available forum discussions on computing the equivalent number of looks (i.e. Equivalent number of looks (ENL)) and see that in the current SNAP version the ENL is computed within the Analysis/Statistics tool.

I don’t perform multilooking on my RADARSAT-2 quad-pol, and instead use the Refined Lee speckle filter to retain the Intensity bands. I need to estimate the equivalent number of looks at the end of my processing. My subset does not offer much in terms of homogenous areas; they roughly contain 100 000 pixels and are not super homogeneous. The resulting ENL varies between 8 and 0.7 depending on the image and area selected.

What should I do? Is it just impossible to estimate ENL in my case?

Thank you for your insight!


would it make sense to estimate the ENL before you create a subset? You would have larger homogenous areas left and the degree of speckle should be consistent within the entire image (at least with respect to wavelength and spatial resolution, not necessarily for different surfaces).

Thank you for your advice. Here is the most homogenous area I could find in all of my unsubsetted images. It contains 155 000 pixels. Before speckle filtering (top), its ENL is 0.92 (shouldn’t it be 1?). After the Refined Lee (bottom), ENL for the same area is 5.6. Do you think this number is representative? And for similar images processed in the same way, should the ENL be roughly the same?

I have trouble trusting these results since I don’t even get an ENL value of 1 for the un-filtered image…

maybe that’t a bit too large. Depending on the resolution and wavelength, some speckle filters can even increase unwanted patterns, as in this case. How about one of these dark patches?