Tracking Image Changes with IMAGEJ

[Catherine Zucker]

Hi,
I'm a high school senior researching dune migration on Mars. I need to measure how many meters the dunes have moved in a certain number of months. I've been trying to do this with IMAGEJ, but am finding it difficult. Does anyone know a way to track the migration of an entire dune using IMAGEJ? Does anyone have experience with tracking image changes? Thanks!

[dpoburko]

Hi Catherine,

  Sounds like a fun project. We could probably use a little more info on
what kinds of features you want to track. Do the dunes stand out clearly
from other features? Do they have a common shape? Are you working with
color or grayscale images? How do you define "dune" vs "not dune" in the
images? It might be helpfull if you could provide some example imags.
I'm sure ImageJ is capable of measuring what you want to know. Just be
sure to provide a little more detail on what exactly you want to
measure. Also, how many images are you planning to ananlyze. Is it few
enough that it will be reasonable to do a few steps by hand?

Cheers,
Damon

On 12/14/2010 10:08 AM, Catherine Zucker wrote:
> Hi,
> I'm a high school senior researching dune migration on Mars. I need to measure how many meters the dunes have moved in a certain number of months. I've been trying to do this with IMAGEJ, but am finding it difficult. Does anyone know a way to track the migration of an entire dune using IMAGEJ? Does anyone have experience with tracking image changes? Thanks!

[Gabriel Landini]

On Tuesday 14 December 2010 13:08:17 Catherine Zucker wrote:
> I'm a high school senior researching dune migration on Mars. I need to
> measure how many meters the dunes have moved in a certain number of months.
> I've been trying to do this with IMAGEJ, but am finding it difficult. Does
> anyone know a way to track the migration of an entire dune using IMAGEJ?
> Does anyone have experience with tracking image changes? Thanks!

I have no idea of what the images look like, but I guess that you will have
some lines that delineate the dunes?
If so, you could isolate the area that the lines define and use some kind of
distance transform between these. You will need to calibrate the images and
make sure that the 2 images are aligned (registered) on some other fixed
landmarks.

Good luck with your project, I wished there were projects like these when I
was in high school.

Cheers

Gabriel

[D Hill]

Well I would guess that you would want to look for an edge tracking tool
(fourier transforms), and use the top edge of a dune as a reference. Then I
guess it would be a matter of using trigonometry to calculate distances

Duncan

On Wed, Dec 15, 2010 at 9:54 AM, Gabriel Landini <[hidden email]>wrote:

> On Tuesday 14 December 2010 13:08:17 Catherine Zucker wrote:
> > I'm a high school senior researching dune migration on Mars. I need to
> > measure how many meters the dunes have moved in a certain number of
> months.
> > I've been trying to do this with IMAGEJ, but am finding it difficult.
> Does
> > anyone know a way to track the migration of an entire dune using IMAGEJ?
> > Does anyone have experience with tracking image changes? Thanks!
>
> I have no idea of what the images look like, but I guess that you will have
> some lines that delineate the dunes?
> If so, you could isolate the area that the lines define and use some kind
> of
> distance transform between these. You will need to calibrate the images and
> make sure that the 2 images are aligned (registered) on some other fixed
> landmarks.
>
> Good luck with your project, I wished there were projects like these when I
> was in high school.
>
> Cheers
>
> Gabriel
>

[Michael Doube]

Hi Catherine

If you have a sequence of images taken from a static viewpoint you can subtract the first image from the rest of the images and this will give a rough idea of how different the subsequent images are.  This is easily achieved in ImageJ by opening an image sequence as a stack, duplicating the first image, then using the image calculator to subtract the duplicated first image from all the images in the stack.

Would be interesting to see your results!

Michael

On Tuesday 14 December 2010 13:08:17 Catherine Zucker wrote:
> I'm a high school senior researching dune migration on Mars. I need to
> measure how many meters the dunes have moved in a certain number of months.
> I've been trying to do this with IMAGEJ, but am finding it difficult. Does
> anyone know a way to track the migration of an entire dune using IMAGEJ?
> Does anyone have experience with tracking image changes? Thanks!

[Catherine Zucker]

Hi! I asked a question about tracking image changes a while back. Thanks for
all the help! I was wondering if anyone could give me more specific advice
if I attached the actual images I was using. All my images are grayscale.
They all come off of the HIRISE imaging database, which takes high
resolution images of the Martian surface. The dunes are pretty clearly
delineated via their raised ridges. The files are too large to attach, but
I've included the links to two images I’m analyzing from the HIRISE
database. I'm working with the images entitled "full image (grayscale, map
projected)" found in the link on the right hand side of the page. Both
images have a scale of 50 cm/pixel. In the bottom right corner of each
image, inside the impact crater, there's a raised sand mound that I know
hasn't moved. Specifically, I'm trying to measure how many meters the sand
dunes above that impact crater have moved from one image to the next. It
doesn't need to be perfect. I just need a rough estimate. At most, I'll
analyze two or three image sets. I can do as many steps by hand as is
necessary to achieve a successful result. Thank you so much! ~Catherine


From November 2007:
http://hirise.lpl.arizona.edu/PSP_005931_1080
From July 2009:
http://hirise.lpl.arizona.edu/ESP_013829_1080


On Tue, Dec 14, 2010 at 1:56 PM, Damon Poburko <[hidden email]>wrote:

> Hi Catherine,
>
>  Sounds like a fun project. We could probably use a little more info on
> what kinds of features you want to track. Do the dunes stand out clearly
> from other features? Do they have a common shape? Are you working with color
> or grayscale images? How do you define "dune" vs "not dune" in the images?
> It might be helpfull if you could provide some example imags. I'm sure
> ImageJ is capable of measuring what you want to know. Just be sure to
> provide a little more detail on what exactly you want to measure. Also, how
> many images are you planning to ananlyze. Is it few enough that it will be
> reasonable to do a few steps by hand?
>
> Cheers,
> Damon
>
>
> On 12/14/2010 10:08 AM, Catherine Zucker wrote:
>
>> Hi,
>> I'm a high school senior researching dune migration on Mars. I need to
>> measure how many meters the dunes have moved in a certain number of months.
>> I've been trying to do this with IMAGEJ, but am finding it difficult. Does
>> anyone know a way to track the migration of an entire dune using IMAGEJ?
>> Does anyone have experience with tracking image changes? Thanks!
>>
>

[Michael Schmid]

Hi Catherine,

it looks problematic - the elevation of the sun was different in the  
two images, and this may lead to an apparent shift of the dunes  
(unless the crests are sharp enough so that the sunlight-shadow  
border is independent of sun elevation).

Roughly, you could try the following steps:
- Align the images such that a known stable feature is at the same  
pixel position in both (e.g., measure the position of that feature  
with the cursor, translate one of the images until it matches.)
The Sync Windows plugin from the ImageJ website may help.
- Identify the crests (e.g. Process>Filters>Convolve with a kernel  
like "-1 0 1" or reverse for crests running in y direction), and try  
various filters to make them stand out clearly, suppressing all other  
signal as much as possible.
- You may then see how much you have to shift the images to make the  
dunes conincide (e.g. copy, paste with "Subtract" in paste Control,  
and nudge with the cursors) or try the bUnwarpJ plugin on the two  
images to get the shift.


Michael
________________________________________________________________

On 2 Jan 2011, at 22:40, ImageJ Lover wrote:

> Hi! I asked a question about tracking image changes a while back.  
> Thanks for
> all the help! I was wondering if anyone could give me more specific  
> advice
> if I attached the actual images I was using. All my images are  
> grayscale.
> They all come off of the HIRISE imaging database, which takes high
> resolution images of the Martian surface. The dunes are pretty clearly
> delineated via their raised ridges. The files are too large to  
> attach, but
> I've included the links to two images I’m analyzing from the HIRISE
> database. I'm working with the images entitled "full image  
> (grayscale, map
> projected)" found in the link on the right hand side of the page. Both
> images have a scale of 50 cm/pixel. In the bottom right corner of each
> image, inside the impact crater, there's a raised sand mound that I  
> know
> hasn't moved. Specifically, I'm trying to measure how many meters  
> the sand
> dunes above that impact crater have moved from one image to the  
> next. It
> doesn't need to be perfect. I just need a rough estimate. At most,  
> I'll
> analyze two or three image sets. I can do as many steps by hand as is
> necessary to achieve a successful result. Thank you so much!  
> ~Catherine
>
>
> From November 2007:
> http://hirise.lpl.arizona.edu/PSP_005931_1080
> From July 2009:
> http://hirise.lpl.arizona.edu/ESP_013829_1080
>
>
> On Tue, Dec 14, 2010 at 1:56 PM, Damon Poburko  
> <[hidden email]>wrote:
>
>> Hi Catherine,
>>
>>  Sounds like a fun project. We could probably use a little more  
>> info on
>> what kinds of features you want to track. Do the dunes stand out  
>> clearly
>> from other features? Do they have a common shape? Are you working  
>> with color
>> or grayscale images? How do you define "dune" vs "not dune" in the  
>> images?
>> It might be helpfull if you could provide some example imags. I'm  
>> sure
>> ImageJ is capable of measuring what you want to know. Just be sure to
>> provide a little more detail on what exactly you want to measure.  
>> Also, how
>> many images are you planning to ananlyze. Is it few enough that it  
>> will be
>> reasonable to do a few steps by hand?
>>
>> Cheers,
>> Damon
>>
>>
>> On 12/14/2010 10:08 AM, Catherine Zucker wrote:
>>
>>> Hi,
>>> I'm a high school senior researching dune migration on Mars. I  
>>> need to
>>> measure how many meters the dunes have moved in a certain number  
>>> of months.
>>> I've been trying to do this with IMAGEJ, but am finding it  
>>> difficult. Does
>>> anyone know a way to track the migration of an entire dune using  
>>> IMAGEJ?
>>> Does anyone have experience with tracking image changes? Thanks!
>>>
>>

[Stephan Saalfeld]

Hi Catherine,

I've played with the images in Fiji for some minutes now and I got the
following results/problems:

The map-projected images seem to not be very precisely projected.  I
reason that from seeing both the dunes and the immobile sand mount move
into the same direction.

Assuming that at least the rotational component is correct, I set a
point (+) on the tip of the sand mount in each image and used the plugin

http://pacific.mpi-cbg.de/wiki/index.php/Landmark_Correspondences

using `Translation' to shift one image into the other.  If you know
about more points that do not move, you could use higher order
transformations to align the images such as

* rigid (shift and rotation, >= 2 points)
* similarity (shift, rotation, isotropic scale, >= 2 points)
* affine (shift, rotation, nonisotropic scale, shear, >= 3 points)
* perspective (affine + perspective projection, >= 4 points)
* moving least squares free deformation with one of the above models
  used for interpolation if the deformation

I see a problem here because the images mainly consist of dunes that are
expected to move such that they cannot be used for alignment.  The task
looks a bit ill posed here.  Otherwise, automatic image based alignment
would be trivial:

http://pacific.mpi-cbg.de/wiki/index.php/Feature_Extraction

and, again

http://pacific.mpi-cbg.de/wiki/index.php/Landmark_Correspondences

Assuming that you have the images aligned, you should have a look at the
plugin

http://pacific.mpi-cbg.de/wiki/index.php/PIV_analyser

or, alternatively, an unpublished plugin from ourselves that was a
predecessor of the above and hacked back in 2008:

http://fly.mpi-cbg.de/saalfeld/download/Optic_Flow.class
http://fly.mpi-cbg.de/saalfeld/download/Optic_Flow.java

Both plugins measure, for each pixel in two adjacent images of a
time series, where it has moved using an image based similarity measure
for the local vicinity of the pixel.

The plugins differ in some details, PIV_analyser gives you more precise
results to the cost of much higher runtime.  The differences are mainly:

                    PIV_analyser         Optic_Flow
local vicinity      square block         Gaussian bell shape
similarity measure  Pearson correlation  sum of differences
speed               slow                 fast

With the sand mount single point alignment mentioned above, both plugins
show that the upper part moves faster than the lower part of the image.
The simple explanation for that is that the images are still not aligned
precisely in terms of rotation.  Do you know about the distance of the
camera to the surface?  If not, there might also be a scale factor to
compensate for...

However, good luck :)

Best regards,
Stephan
 
--




On Sun, 2011-01-02 at 16:40 -0500, ImageJ Lover wrote:

> Hi! I asked a question about tracking image changes a while back. Thanks for
> all the help! I was wondering if anyone could give me more specific advice
> if I attached the actual images I was using. All my images are grayscale.
> They all come off of the HIRISE imaging database, which takes high
> resolution images of the Martian surface. The dunes are pretty clearly
> delineated via their raised ridges. The files are too large to attach, but
> I've included the links to two images I’m analyzing from the HIRISE
> database. I'm working with the images entitled "full image (grayscale, map
> projected)" found in the link on the right hand side of the page. Both
> images have a scale of 50 cm/pixel. In the bottom right corner of each
> image, inside the impact crater, there's a raised sand mound that I know
> hasn't moved. Specifically, I'm trying to measure how many meters the sand
> dunes above that impact crater have moved from one image to the next. It
> doesn't need to be perfect. I just need a rough estimate. At most, I'll
> analyze two or three image sets. I can do as many steps by hand as is
> necessary to achieve a successful result. Thank you so much! ~Catherine
>
>
> From November 2007:
> http://hirise.lpl.arizona.edu/PSP_005931_1080
> From July 2009:
> http://hirise.lpl.arizona.edu/ESP_013829_1080
>
>
> On Tue, Dec 14, 2010 at 1:56 PM, Damon Poburko <[hidden email]>wrote:
>
> > Hi Catherine,
> >
> >  Sounds like a fun project. We could probably use a little more info on
> > what kinds of features you want to track. Do the dunes stand out clearly
> > from other features? Do they have a common shape? Are you working with color
> > or grayscale images? How do you define "dune" vs "not dune" in the images?
> > It might be helpfull if you could provide some example imags. I'm sure
> > ImageJ is capable of measuring what you want to know. Just be sure to
> > provide a little more detail on what exactly you want to measure. Also, how
> > many images are you planning to ananlyze. Is it few enough that it will be
> > reasonable to do a few steps by hand?
> >
> > Cheers,
> > Damon
> >
> >
> > On 12/14/2010 10:08 AM, Catherine Zucker wrote:
> >
> >> Hi,
> >> I'm a high school senior researching dune migration on Mars. I need to
> >> measure how many meters the dunes have moved in a certain number of months.
> >> I've been trying to do this with IMAGEJ, but am finding it difficult. Does
> >> anyone know a way to track the migration of an entire dune using IMAGEJ?
> >> Does anyone have experience with tracking image changes? Thanks!
> >>
> >

[Michael P Ellis]

On my MacBookPro running OS 10.6.7 with java version "1.6.0_22" the macro recording window appears to no longer display a macro as it is being recorded.


Reproduce Instructions
        Open ImageJ
        Select Plugins->Macros->Record...
        // The Macro recorder window appears but the text area where the macro used to be displayed is absent
        // the controls for creating and naming the macro are displayed. You can grow the window to display the
        // text area by dragging the corner of the window

        Select File->New->Image... and create a new window
        // This ought to display something like
        //          newImage("Untitled", "8-bit Ramp", 400, 400, 1);
        // Nothing is displayed in the text area

        Click on the "Create" button in the "Recorder" window
        // A macro window with the commands appears (apparently out of nowhere)


This behaviour I have reproduced with ImageJ 1.440 (64 bit and 32 bit) and many of the older versions of ImageJ too so I suspect this is down to a change in some configurational change on my Mac (probably java) on OS-X.

Anyone else seen this problem and or know of a solution?

-- Michael Ellis

[Glen MacDonald-2]

Hi,
the macro recorder works as expected with my MBP, OS 10.6.7, same java, IJ 1.45e10.  Maybe try reinstalling Imagej?

On the other hand, the 3D Viewer plugin displays a greatly zoomed image that is unusable on this MBP.  The same plugin works just fine everywhere else.  

Regards.
Glen
Glen MacDonald
Core for Communication Research
Virginia Merrill Bloedel Hearing Research Center
Box 357923
University of Washington
Seattle, WA 98195-7923  USA
[phone]
[hidden email]








On Mar 31, 2011, at 5:16 AM, Michael Ellis wrote:

> On my MacBookPro running OS 10.6.7 with java version "1.6.0_22" the macro recording window appears to no longer display a macro as it is being recorded.
>
>
> Reproduce Instructions
> Open ImageJ
> Select Plugins->Macros->Record...
> // The Macro recorder window appears but the text area where the macro used to be displayed is absent
> // the controls for creating and naming the macro are displayed. You can grow the window to display the
> // text area by dragging the corner of the window
>
> Select File->New->Image... and create a new window
> // This ought to display something like
> //          newImage("Untitled", "8-bit Ramp", 400, 400, 1);
> // Nothing is displayed in the text area
>
> Click on the "Create" button in the "Recorder" window
> // A macro window with the commands appears (apparently out of nowhere)
>
>
> This behaviour I have reproduced with ImageJ 1.440 (64 bit and 32 bit) and many of the older versions of ImageJ too so I suspect this is down to a change in some configurational change on my Mac (probably java) on OS-X.
>
> Anyone else seen this problem and or know of a solution?
>
> -- Michael Ellis