HICO LSI november

TCC
logarithmic enhancement

FCC
logarithmic enhancement

TCC deglinted
same logarithmic enhancement
"Glint" actually is caused by variable atmospheric path

FCC deglinted
same logarithmic enhancement
"Deglinting" efficiently removes variable atmospheric path,
thanks to excellent correlations between NIR band and all other HICO bands:
see glint regressions text file

lsihico_20101123
The AutoCalibration process

Special effect mask
This is prepared automatically.

Note that threshold on deglinting
leaves some critical area not-deglinted;
therefore, threshold should be upped a bit.

Lands and Clouds
are mapped automatically
by the AutoCalibration process,
with some manual editing
through shapefiles.

Image preparation
and AutoCalibration
is comprised of the following steps:

create the working database.pix in PCIDSK format
- 106 channels
- complete with georeferencing
- see layout of database.pix
import raw data in 16U channels
create a synthetic Panchromatic band and writes it as band 27 in database.pix
scale 28 raw 16U data bands to 28 raw 8U range
use various shapefiles to prepare a special effect mask in channel 29 (see left)
estimate spectral LsM, Lsw, and some other stuff
estimate glint regressions
extract deglinted calibration data
- see lsihico_20101123_m0.cal
format and display a AutoCalibration diagram and display it
format a Autocal.txt textfile which contains a proposed command line, ready for fine-tuning

The AutoCalibration diagram

The proposed AutoCalibration command line
now needs attention:
changes are to be fine-tuned to suit
the real life image calibration case.

lsihico_20101123
The calibration process

Calibration for bands
blue=7, green=13, red=22 and NIR=28

Calibration for bands
blue=7, 12, 21 and PAN

All combinations of bands must be investigated,
in order to fine tune
the simplified radiance transfer model

In 4SM,
other than Jerlov's table
of diffuse attenuation coefficients for irradiance 400-700 nm
for marine waters,
no "field" data is used
in the process of calibrating
any shallow scene

My experience is that wavelength for the Blue and Green bands are to be left unchanged, so that Jerlov's dataset is interpolated using the observed K_blue/K_green=0.60 ratio of diffuse attenuation coefficients with WLblue=457.7 nm and WLgreen=544.4 nm
Except of course for the PAN band, all other wavelengths are to be fine tuned, so as to obtain the best possible fit for all pairs of bands
- this involves significant decrease of effective 2K values, in the near blue range, in the far green range, and in the whole red range
- see Jerlov's quotation
As for the PAN band, it is modelled using the optical properties observed for all other wavebands involved

lsihico_20101123
Deglinting along Profile Red

lsihico_20101123_m0.sig textfile

Deglinting along Profile Red

No smoothing applied: this shows the real amount of noise
See Profile Red locator

LOWER: from waypoint 1 to waypoint 2, the profile runs over a very dark-bottomed seagrass bed
UPPER: from waypoint 5 onwards, the profile runs over optically deep waters

I have come to suspect that deep water radiances for bands in the red range are significantly higher over shallow areas than over eastern oceanic waters