Analyzing extremely large images in ImageJ

[Trevor Steve]

Hi All,

I am wondering if anyone has developed a work-around to enable analysis of
extremely large images in ImageJ.  I am working with such images:
E.G. 182612 x 215608 (> 39 GPIX) which are obviously too large for the
2GPIX ImageJ plane.

I wonder if anyone has developed a way to process such images "in tiles
using XY offsets".  It has been suggested that one could "write a macro
that loops over portions of the image plane to perform processing that
way", but this is far beyond my skill-set.

Thanks for any suggestions,

Trevor

--
Trevor Steve, MD

Neurology Resident (PGY-7)
PhD Student, Department of Medicine
University of Alberta

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Majurski, Michael Paul (Fed)]

When performing analysis on images of this size the type of operation controls the difficulty. Things like pixel-wise operations are simple enough, however more complex operations like convolution can quickly become very messy to handle because of managing overlap regions. As far as I am aware there is no way to perform ImageJ operations on a disk backed image currently. Libvips does exist for doing this sort of thing, however it is written in C and C++ with a Python wrapper as opposed to bundled in something like ImageJ.

Interestingly enough I am working on something in this area but it is still very much in the design phase.

Would you be willing to share the details of your problem? Specifically how the images are currently manipulated and stored? If so you can reply to me directly to avoid cluttering the mailing list.

Regards,
Michael Majurski
[hidden email]

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Trevor Steve
Sent: Tuesday, September 09, 2014 12:35 PM
To: [hidden email]
Subject: Analyzing extremely large images in ImageJ

Hi All,

I am wondering if anyone has developed a work-around to enable analysis of extremely large images in ImageJ.  I am working with such images:
E.G. 182612 x 215608 (> 39 GPIX) which are obviously too large for the 2GPIX ImageJ plane.

I wonder if anyone has developed a way to process such images "in tiles using XY offsets".  It has been suggested that one could "write a macro that loops over portions of the image plane to perform processing that way", but this is far beyond my skill-set.

Thanks for any suggestions,

Trevor

--
Trevor Steve, MD

Neurology Resident (PGY-7)
PhD Student, Department of Medicine
University of Alberta

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[ctrueden]

Hi Michael,

> As far as I am aware there is no way to perform ImageJ operations on a
> disk backed image currently.

One of the goals of ImageJ2 is to provide this feature. However, the
current version can only accomplish it in some scenarios.

In a nutshell: extremely large _numbers_ of image planes are transparently
cached to and from disk by the SCIFIO library, such that you can make
changes to a "virtual stack" and have them survive when the position
sliders are moved back and forth.

Unfortunately, extremely large image planes such as those in Trevor's
datasets are not yet cached tilewise. It is a future goal of the project,
though:

* https://github.com/scifio/scifio/issues/125
* https://github.com/imagej/imagej/issues/87

Regards,
Curtis

On Wed, Sep 10, 2014 at 9:29 AM, Majurski, Michael Paul <
[hidden email]> wrote:

> When performing analysis on images of this size the type of operation
> controls the difficulty. Things like pixel-wise operations are simple
> enough, however more complex operations like convolution can quickly become
> very messy to handle because of managing overlap regions. As far as I am
> aware there is no way to perform ImageJ operations on a disk backed image
> currently. Libvips does exist for doing this sort of thing, however it is
> written in C and C++ with a Python wrapper as opposed to bundled in
> something like ImageJ.
>
> Interestingly enough I am working on something in this area but it is
> still very much in the design phase.
>
> Would you be willing to share the details of your problem? Specifically
> how the images are currently manipulated and stored? If so you can reply to
> me directly to avoid cluttering the mailing list.
>
> Regards,
> Michael Majurski
> [hidden email]
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
> Trevor Steve
> Sent: Tuesday, September 09, 2014 12:35 PM
> To: [hidden email]
> Subject: Analyzing extremely large images in ImageJ
>
> Hi All,
>
> I am wondering if anyone has developed a work-around to enable analysis of
> extremely large images in ImageJ.  I am working with such images:
> E.G. 182612 x 215608 (> 39 GPIX) which are obviously too large for the
> 2GPIX ImageJ plane.
>
> I wonder if anyone has developed a way to process such images "in tiles
> using XY offsets".  It has been suggested that one could "write a macro
> that loops over portions of the image plane to perform processing that
> way", but this is far beyond my skill-set.
>
> Thanks for any suggestions,
>
> Trevor
>
> --
> Trevor Steve, MD
>
> Neurology Resident (PGY-7)
> PhD Student, Department of Medicine
> University of Alberta
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html