[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20001024: IMGREMAP: MAG= vs RES= (cont.)

>From: Jason Allard <address@hidden>
>Organization: PSU
>Keywords: 200010240345.e9O3jC425238 McIDAS-X IMGREMAP MAG RES

Jason,

>Thanks for the response... I'm getting a handle on it now.  There are still 
>a couple of questions.

OK.

re: The best way to compare pixels from two images to leave one of the
images in its native projection (since you would then minimize the
massaging of one part of the data being compared) and remap the other
into it.  After doing this, you are guaranteed to have colocated pixel
values that you can then compare.

>Okay, that makes sense.  The reason that I wanted to reproject it is so
>that when I specify the upper left corner and the size, I get the area
>that I want.  If I choose the upper left and 260 by 260 size (on the
>original projection), then the area drawn includes too much area in the
>bottom left and not enough in the bottom right.  The core area of the
>US that I want is the Midwest (MN, IA, MO, WI, IL, IN, MI, OH, KY).
>While I haven't done this yet, if the projection is in lamberts and I
>select the upper left and the size, then I'd get a satellite image that
>is more focused on my area of interest.  Does this make sense to you?

Yes, this makes sense.  Remapping into a rectilinear or mercator
may be better for you in selecting the area you are interested in, but
that is not all that important.

>Basically, when choosing the upper left corner and size 260 by 260, I
>get an image that includes CO, NE, OK and cuts out part of OH and KY.
>I could increase the size to 300 by 300, but my core area would not be
>the focus.

OK.

re: You can also remap one image into a different projection (like lambert
conformal as you mention below) specifying the output RES= resolution.
Instead of remapping into the same projection for the other image,
I would copy the first remapped image and then remap into it.  This
way you are guaranteed of both pixel colocation and pixel magnification.

>So, I would IMGREMAP one the way I want it to be (the right projection,
>the right size, the right resolution, etc), copy that to a new AREA and
>then IMGREMAP the other one to that AREA?

Yes, exactly.

re: My first example had the implicit assumption that the VIS and IR images
either not the same size, or did not have the same registration.
In this case, you _must_ transform (remap) one image into the
projection of the other so that you can do a pixel-by-pixel
comparison.  After all, the objective has to be the comparison of pixel
values at the same geolocation for both images.  This is why I
recommended that you make a copy of the IR image and remap the VIS one
into it.

>This makes sense.  Just to let you know, all the GOES images I have
>where obtained from the SSEC through the deal penn state has with
>Unidata.  So, its the free stuff you offer.  Does that make a
>difference,

No.

>i.e., are the images the same size and have the same registration?

You can test this easily by loading a VIS image in one frame blown down
by a factor of two and the other loaded normally.  Draw maps on both
and then animate the two frames.  Does the map move?  If yes, then
the originals are not registered the exact same way.  Your remapping
will, of course, insure that they resultant pixels are colocated so
you can do comparisons on them.

re: use IMGCHA to set the image header values for remapped imagery
>Oh, okay... it'd best do the IMGCHA command.

Yes.  You will want to use the images again and again most likely, so
it will be most useful to know what they really are.

re: Right, so the best thing to do is remap one into a copy of the other
and then compare the remapped copy with the original.  This was the
example I sent you.

>Okay, I get this in theory, and I think I understand the commands, so far

Excellent.

re: Do the images have to be in Lambert Conformal projection as well?  If
not, I would leave the images in their native satellite projection.

>See above for my reasoning of changing to lambert projection.

I understand your objective more clearly now.

>Okay, this makes sense... I think.  Remapping the other image (VIS)
>into the clone (IR) will fit the VIS image to the same location as the
>IR, but will replace the data with the VIS data?

Yes, remapping moves the pixel values from one image to the copy, so
the IR copy that you remap into will end up with VIS data.  This is
what you are after.

re: Use IMGCOPY to subsect and keep the same navigation

>I understand the reason for this, but given that I want to change the
>projection, the best thing is to use the IMGREMAP?

Yes.

re: Will you eventually be doing comparisons between images
from the early 90s to ones from the late 90s?

>I will compare VIS and IR in the early ninties (e.g., VIS May 10, 1991
>at 18 UTM and IR May 10, 1991 at 18 UTM) and VIS and IR in the late 90s
>(e.g., VIS June 20, 1998 at 18 UTM and IR June 20, 199 at 18 UTM).  The
>only reason I'll compare early 90s to late 90s will be after I've
>determined the frequency of cumulus (to see if there are similar
>frequencies or if there are different areas that have higher cumulus
>frequencies).

OK, I just wanted to make sure...

re: By doing this, you are going to modify the data values in both images
and then compare the modified values.  As long as you realize this,
you will not get into interpretation difficulties.

>I'm not completely sure what you mean.  How significant are the
>modifications?

The greater the difference in original projection and the final projection,
the greater effective change in the remapped values.

>Why are they modified?

If the input and output resolutions are matched, then the differences
are minimized.  Think about what is going on:  in the original image
there is a certain spacing between pixels; when you remap to a new
projection, the spacing has to change.  In order to fill the resultant
values at each pixel location, averages must be made from sorrounding
input pixels to get output values.

Consider the following, very crude example:


        original projection      remapped projection

        ....................        ..........
         ..................        ............
          ................        ..............
           ..............   ->   ................
            ............        ..................
             ..........        ....................
              ........        ......................


How can you transform the "image" on the left hand side to the "image" on
the right hand side and not lose data points near the top and create new
data points on the bottom?

re:  IMGREMAP WF/051091IR.19 WF/SUB051091IR.19 LAT=55 100 PLA=CENTER PRO=LAMB 
SIZ=260 260 MAG=1 RES=8
     IMGCOPY WF/SUB051091IR.19 MYDATA/IMAGES.1234 SIZE=ALL
     IMGREMAP WF/051091VIS.8 MYDATA/IMAGES.1234>   

>Okay, this makes sense.  IMGREMAP'ing the VIS into the modified IR
>makes all the changes to the VIS that I wanted... i.e., changing the
>resolution, the size, the projection.  Correct?

Right!

re: Now, the images that are comparable are WF/SUB051091IR.19 and
MYDATA/IMAGES.1234.

>Okay, that's what I wanted.

re: I always recommend, however, to do as
little data massaging as possible.  This is why I immediately went into
not modifying one image while remapping the other into a copy of the
first.

>I understand... that would be ideal, but without changing the
>projection, I'm not sure if the area that I select with choosing the
>upper left and the size will be the area that I want.

OK.

>This has really helped... assuming what I've written above is valid,
>then I think that I'm getting a handle on all this.

Yes, I think you are catching on nicely.

>Thanks,

You are welcome.

Tom