4SM illustrations the_deglinting

"Do you know why deglinting works well in some cases, poorly in others?"++++++++++
- You need a very good correlation, with R2~=1, see https://www.watercolumncorrection.com/web-gezirat.php#deglinting
- For lousy and complex/noisy correlations, see

OVERCAST+++++++++++++++++
- I suspect that
  - * slightly overcast skies is best (whitish sky, like in the MiddleEast)
  - * very clear deep-blue sky is worst (you get something like rainbowed discoloration) like in your Kauai image

WHITECAPS+++++++++++++
- I'm not sure about whitecaps and boat wakes

CLOUDS/HAZE/ADJACENCY_EFFECT+++++++++++++
- Clouds and haze,
- and adjacency effect underr overcast skies,
- would appear to have spectral properties similar to skyglint

OBLIQUE+++++++++++++++
- to minimize glint, I want to view from Nadir to away from the sun,
- NOT into the sun!!!!!!!!!

Time differences between waveband acquisition++++++++++++++
- this can explain why Glint correlation decreases for two bands farther apart
- see https://www.watercolumncorrection.com/web-gezirat.php#deglinting


	Cloud and its shadow at Caicos Bank Inter-band correlation is very strong.
	Glint and Cloud at Tarawa part of the cloud has been added to the scatter plot Sea-surface glint, and haze that surrounds the cloud, seem to exhibit very similar correlation here, as they both mostly tell us about Mie scattering in the atmosphere. They both may be removed through the same "deglinting" process.
	Sea-surface glint is fairly strong in this wind-swept SPOT image Inter-band correlation is reasonable. Smoothing can greatly improve this correlation. Deglinting at Tarawa removes specular reflection at the sea surface : the sky glint from the sky dome the sun glint from direct sun light isolated lumps of haze, and the haze that surrounds the clouds
	<== Here a rectangular window has been left untouched in order to illustrate the -E extraction window feature of 4SM. Deglinting blandly transfers the noise from the NIR band into other bands!

	Upper : modeling on raw data note the presence of a faint lump of haze and the corresponding artifact in the form of underestimated depth and bottom brightness Deglinting at Caicos Bank Lower : modeling on deglinted data see how deglinting has removed the haze (while its shadow is still there) and the corresponding improvement in the final result


	Deglinting at Lee Stocking Island, Bahamas (Ikonos) CENTRAL : raw data BACKDROP : smoothed-deglinted data

	Deglinting at MahoneBay, Nova Scottia, Canada (CASI) Sunglint affect this area. poor correlation between the Nir band ond other bands deglinting based on the red bands over this fairly deep area. Removing the sunglint also removes the skyglint. The profile compares RAW in blue DEGLINTED in red Smart-Smoothed in green FINAL in black
	Smart-Smoothing preserves the edges at MahoneBay, Nova Scottia, Canada (CASI) No sunglint here Smart-Smoothing preserves the edges of the dark and shallower patch The profile compares RAW in blue DEGLINTED in red Smart-Smoothed in green FINAL in black

	Deglinting of a CASI mosaic at Gezirat Siyul Reef, Red Sea

This image preparation is done
"on the fly" while modeling,
as 4SM reads the raw U8 or U16 data
and applies all the required improvements
on a pixel wise basis prior to modeling.

Radiometric improvements:
The Delimbing/Deglinting process
on a 512 pixels wide raw CASI flight line

from bottom to top:

Raw data: the limb brightening is very obvious: it must be removed prior to geocoding of the image
Delimbed/Derowed: note that in spite of the derowing of individual rows, there remains groups of adjacent rows which are slightly brighter or darker: this shows how important for shallow water work that the imaging instrument be optimally tuned
Delimbed/Derowed/Deglinted/Smart-smoothed
Delimbed/Derowed/Deglinted/NOT-smoothed: in spite of all efforts, there remains all sorts of noise features and also remnants of imperfect deglinting: therefore it is very important that the data be smoothed prior to modeling