The differences between the sensitivity of the human eye and onboard
satellite instruments is quite striking with MSG-1 at the moment. Current
visual imagery of Europe obviously shows low levels of illumination - and
indeed darkness north of ... whatever the current latitude is.
Yet here in N. Wales, even under quite heavy cloud, everything is bright
(if not particularly cheerful). Our eyes seem much better at responding to
lower levels of light than the instruments. On the other hand, we aren't very
good at seeing things by infrared!
If I haven't got this all wrong then a question follows: Are there any ways
in which we might use our software to squeeze a little bit more out from
the visible wavelengths at the darkest parts of the year? I'm not at all clear
about this. I appreciate, of course, that one can't enhance something that is
not there. David (T) and others, do you have any thoughts on this, or am I
wasting everyone's time with the question?
Robert



------------------------------
Robert Moore

tel and fax: 44 (0) 1352 714456

i don't know if i understood your question correctly, but i think histogram equalization of the visible images could solve your problem.

In a message dated 20-11-03 16:10:38 GMT Standard Time,
rsmoore@liverpool.ac.uk writes:

Yes I was thinking how I can see as easily at summer noon and after dark in
winter with a full moon. The latter level of illumination is a minute
fraction
of the former - yet no MSG vis imagery by moonlight!

Robert,

No, MSG will never image by moonlight but at last years EUMETSAT conference
there were serious suggestions about doing just that in the next generation
of
geostationary satellites.

Trouble is that imaging via moonlight requires an enormous aperture
radiometer -
which in turn requires an enormous satellite.

Regards,
John Tellick.
RIG.

Robert,

I agree with you that much of the information that may be used to visualise
aspects that are there are lost in the visible light channels. For MSG this is
the current situation and for HRPT it has been. We resolved it for HRPT quite
some time ago by developing an algorithm that deals with both visible and
infrared information. It might be a thought for David to consider incorporating
the same principle for MSG as well.
David, what do you think?

Ferdinand

Here are my two pennies:

MSG (SEVERI) does send 10 bits grey values (0…1023). Not all these
values are used however.
Because of the JPG limitations MDM scales these down to 256. Applying
a gamma on a low light image does give poor results.
But under MDM's advanced setup you can change the min and max value
for every channel separate.
This way the 256 grey values are better distributed and gamma can be
applied, depending on the season.
Using XRIT2PIC (which can handle the full 10 bits) I determined the
most used Min and Max value for every channel. By setting these in
Advanced setup, MDM does utilise the available spectrum much better.
Here are my values:
Ch01=MAX=891,MIN=43,GAM=140
Ch02=MAX=806,MIN=60,GAM=145
Ch03=MAX=710,MIN=43,GAM=100
Ch04=MAX=756,MIN=53,GAM=100
Ch05=MAX=742,MIN=86,GAM=100
Ch06=MAX=645,MIN=74,GAM=100
Ch07=MAX=829,MIN=74,GAM=100
Ch08=MAX=801,MIN=150,GAM=100
Ch09=MAX=799,MIN=88,GAM=100
Ch10=MAX=800,MIN=102,GAM=100
Ch11=MAX=787,MIN=150,GAM=100
Ch12=MAX=822,MIN=68,GAM=135

By the way, automatic equalisation is nice on HRPT images, but can be
annoying on MSG animations !

Greetings,
Arne van Belle

--- In MSG-1@yahoogroups.com, "a_van_belle" <a.van.belle@h...> wrote:

Here are my two pennies:

MSG (SEVERI) does send 10 bits grey values (0…1023). Not all these
values are used however.

The minimum value for any channel on MSG-1 is currently 51 (0..1023)

Because of the JPG limitations MDM scales these down to 256.

If there was enough interest, the full ten bits could be delivered
in the Professional version of the MSG Data Manager. Most present-
day image formats and processing tools can only handle 8-bit images,
indeed they eye can only see about 6 bits.

Applying

a gamma on a low light image does give poor results.
But under MDM's advanced setup you can change the min and max

value for every channel separate. This way the 256 grey values are
better distributed and gamma can be applied, depending on the season.

Seasonal adjustments are an excellent idea! The default values of
gamma and minimum level are carefully chosen so that black is black,
and a single level change above black in the 0..1023 domain will be
reflected as a single level change in the 8-bit final image.

[]

Here are my values:
Ch01=MAX=891,MIN=43,GAM=140
Ch02=MAX=806,MIN=60,GAM=145
Ch03=MAX=710,MIN=43,GAM=100

[]

Greetings,
Arne van Belle

It's great to hear the results of your experiments, but why do you
set the black level on these channels different to what is
transmitted (i.e. 51)?

A word of warning to anyone making such adjustments in to ensure
that your monitor is correctly - see:

http://www.jasc.com/monitor1.asp

Cheers,
David

Our eyes seem much better at responding to lower levels of light

than the instruments.

Robert,

Our eyes (and ears) have a log response, the scientific instruments
on MSG-1 demand a linear response. If Log (signal) were transmitted
over the link all your prayers would be answered! Ah, assuming the
data source has a low noise and a precisely-set black level, that is!

Cheers,
David

Yes I was thinking how I can see as easily at summer noon and after dark in
winter with a full moon. The latter level of illumination is a minute fraction
of the former - yet no MSG vis imagery by moonlight!

Robert

Quoting David Taylor <david-taylor@blueyonder.co.uk>:

Our eyes seem much better at responding to lower levels of light

than the instruments.

Robert,

Our eyes (and ears) have a log response, the scientific instruments
on MSG-1 demand a linear response. If Log (signal) were transmitted
over the link all your prayers would be answered! Ah, assuming the
data source has a low noise and a precisely-set black level, that is!

Cheers,
David



To unsubscribe from this group, send an email to:
MSG-1-unsubscribe@yahoogroups.com



Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

------------------------------
Professor Robert Moore
Department of Sociology, Social Policy
and Social Work Studies
The University of Liverpool
Eleanor Rathbone Building
Bedford Street South
Liverpool
L69 7ZA

tel and fax: 44 (0) 1352 714456

i know that automatic equalization is not reasonable for sequences because of the permanently changing histogramms of the images. that is why i'm doing a 'fixed' histogram equalization. this means i figured out what exactly happens while histogramm equalization of a 12:00 image. this histogram transform is describable by a spline interpolation through a few points (f.e. 20). in this way the resulting image sequences are looking fine. i attached an example of my color composit, where this histogram transform was applied.
max

Also the human eye sees in monochrome at low light levels. Rods and cones !!

Cheers, Guy

--- In MSG-1@yahoogroups.com, "David Taylor" <david-taylor@b...>
wrote:

The minimum value for any channel on MSG-1 is currently 51 (0..1023)

Looking closely at the 10 bit histogram on current HRIT images I
cannot detect this lower limit anymore. (it was there on the initial
image processing system IIPS)

If there was enough interest, the full ten bits could be delivered
in the Professional version of the MSG Data Manager. Most present-
day image formats and processing tools can only handle 8-bit

images,

indeed they eye can only see about 6 bits.

Saving 10 bits would increase the needed storage space too !

It's great to hear the results of your experiments, but why do you
set the black level on these channels different to what is
transmitted (i.e. 51)?

I forgot to mention that I tweaked the ch1,2,3 values a bit to get a
good balance in the R3G2B1 False Colour images.
False Colour is a matter of taste, but I am very pleased with the new
settings (better than the values I used before).

Below are my Min/Max/Gamma values for FSD (except GOES9=8bit)

G10vis=min9,max688,gam135
G10ir=min150,max910,gam100
G10wv=min150,max625,gam100
G12vis=min0,max886,gam135
G12ir=min150,max975,gam100
G12wv=min150,max644,gam100

These values are chosen to maintain all info and spread it as best as
possible into the saved 256 greyscales.
Note that min150 is the Min value upperlimit, not the value I read on
the histogram !

Greetings
Arne van Belle

--- In MSG-1@yahoogroups.com, "a_van_belle" <a.van.belle@h...> wrote:

--- In MSG-1@yahoogroups.com, "David Taylor" <david-taylor@b...>
wrote:

The minimum value for any channel on MSG-1 is currently 51

(0..1023)

Looking closely at the 10 bit histogram on current HRIT images I
cannot detect this lower limit anymore. (it was there on the

initial

image processing system IIPS)

The 51 offset is still there. You will find that the look-at-space
parts of the image are no longer set to 51 (but zero as I recall).
The 51 is still defined in the prolog.

[]

Note that min150 is the Min value upperlimit, not the value I read

on the histogram !

I'll increase the lower limit maximum, then up to 240 or so, and
reduce the upper limit minimum for the next (beta) version.

Cheers,
David

--- In MSG-1@yahoogroups.com, "David Taylor" <david-taylor@b...>
wrote:

I'll increase the lower limit maximum, then up to 240 or so, and
reduce the upper limit minimum for the next (beta) version.

Thanks for that David, I will try this Beta with the ideal values.
Will send you the 10 bits histograms I made later.

Greetings,
Arne van Belle

I'll increase the lower limit maximum, then up to 240 or so, and
reduce the upper limit minimum for the next (beta) version.

Thanks for that David, I will try this Beta with the ideal values.
Will send you the 10 bits histograms I made later.

Greetings,
Arne van Belle

OK, that beta is now up on the Web site.

Cheers,
David