using very large images

[Severino]

I am using some functions of ImageJ for processing earth observation images. ImageJ is so easy to use and so efficient, that it makes an extremely useful tool, even if it has not been developed for remote sensing images and it does not support some functions needed in this area.
As the remote sensing images (from satellites) are extremely large (6000 by 6000 pixels is not one of the largest) I would like to know how I can predict which is the largest image size I can process with ImageJ in my computer. Is there any manual where to find how I could predict it? can it be done at all?
I use frequently TurboReg and I think I could circumvent the size limitation just dividing the large image in smaller ones, performing registration and then merging the registered images again. Has somebody already written some macro script for it? This would be an acceptable solution.
Thank you
Severino

[Mark J. Chopping]

Hi Severino,

You can use the excellent virtual stack feature of ImageJ to avoid size
limitations for most processing jobs, e.g., see the macro in this zip:
http://csam.montclair.edu/~chopping/jornada/EOS/IJ/IJ_LTG7ME-10.zip You
can put the pieces back together again with this trick:
http://csam.montclair.edu/~chopping/jornada/EOS/IJ/makeamontage.html I'm
not sure how to implement your required registration/rectification via
macro but I'm sure someone here will know. Otherwise you are back to
dedicated packages (Imagine, ENVI).

Do you have a new GeoEye-1 image to process by any chance?

  -- Mark

  Mark J. Chopping Ph.D.
  Associate Professor, Earth & Environmental Studies
  Montclair State University, Montclair, NJ 07043
  NASA MISR Science Team, NASA Veg. Structure WG, NACP
  Tel: [phone]  Fax: [phone]
  -------------------------------------------------------
  http://csam.montclair.edu/~chopping/wood

---------

> Date: Fri, 31 Oct 2008 16:37:53 +0100
> From: Severino <[hidden email]>
> Reply-To: ImageJ Interest Group <[hidden email]>
> To: [hidden email]
> Subject: using very large images
>
> I am using some functions of ImageJ for processing earth observation
> images. ImageJ is so easy to use and so efficient, that it makes an
> extremely useful tool, even if it has not been developed for remote
> sensing images and it does not support some functions needed in this
> area. As the remote sensing images (from satellites) are extremely large
> (6000 by 6000 pixels is not one of the largest) I would like to know how
> I can predict which is the largest image size I can process with ImageJ
> in my computer. Is there any manual where to find how I could predict
> it? can it be done at all? I use frequently TurboReg and I think I could
> circumvent the size limitation just dividing the large image in smaller
> ones, performing registration and then merging the registered images
> again. Has somebody already written some macro script for it? This would
> be an acceptable solution. Thank you Severino
>

[Michael Schmid]

Hi Severino,

it depends very much on the type of image (8 bit, 16 bit, 32 bit  
grayscale or 24-bit/pixel RGB) and the operation you want to do.

Most simple operations (3x3 filter operations like smooth, sharpen,  
edge detection) need at least twice the memory needed to store an  
image; for the original and the processed version.

The functions in Process>Filters convert the image to float (32bit/
image) for processing, so you need twice the original image size + 4  
bit/pixel.

Other functions can need even more. I don't know for TurboReg.

ImageJ supports large RAM sizes on some systems, see, e.g. the FAQ  
"How do I increase the memory in ImageJ?" on the documentation wiki
    http://imagejdocu.tudor.lu/

Hope this helps a bit.

Michael
________________________________________________________________

On 31 Oct 2008, at 16:37, Severino wrote:

> I am using some functions of ImageJ for processing earth  
> observation images. ImageJ is so easy to use and so efficient, that  
> it makes an extremely useful tool, even if it has not been  
> developed for remote sensing images and it does not support some  
> functions needed in this area.
> As the remote sensing images (from satellites) are extremely large  
> (6000 by 6000 pixels is not one of the largest) I would like to  
> know how I can predict which is the largest image size I can  
> process with ImageJ in my computer. Is there any manual where to  
> find how I could predict it? can it be done at all?
> I use frequently TurboReg and I think I could circumvent the size  
> limitation just dividing the large image in smaller ones,  
> performing registration and then merging the registered images  
> again. Has somebody already written some macro script for it? This  
> would be an acceptable solution.
> Thank you
> Severino