ImageJ

Image quality evaluation

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

13 messages Options

Emmanuele Sordini

Image quality evaluation

Hello everybody,
I am a new ImageJ user and ML member. My main interest in ImageJ stems from
my amateur astronomy activities (astronomical imaging and image processing).

Let's suppose the following scenario: I have a set of 16-bit B/W frames of
the same target (typically, a planet or a deep-sky object) that have to be
registered, sorted for quality and stacked. The biggest problem I'm faced
with is what image quality indicator should be used for reliable sorting.

The image quality metric should be easy to figure out and not very demanding
in terms of computing power; yet, it should somehow reflect the relative
abundance of details and features on the current frame, which (if I'm not
mistaken) should result in relative abundance of high-frequency components
in the FFT domain. The literature mentions quite a few "one-number"
quantities: RMSE, SNR, peak SNR, and so on... but I'm quite puzzled. I've
also come across some "perceptual indicators" which integrate quantitative
measurement and quality as perceived by the human eye, but I'd rather not
venture into this domain for the moment.

Any ideas/suggestions/links/references on the subject will be highly valued.

Thanks in advance,
Emmanuele Sordini

P.S. Should anybody be interested in my amateur astronomy work, please feel
free to check out my website: www.bloomingstars.com

Thomas Boudier-2

Re: Image quality evaluation

Hi Emmanuele,

As concerns Astronomy i can suggest you to have a look at SalsaJ, a
derivative from ImageJ that is customized for astronomy, it is
multilingual and incorporates special features like aperture intensity
measurement. You can have a look at :
http://www.euhou.net/index.php?option=com_content&task=view&id=7&Itemid=9

As about registration for astronomical images we are currently thinking
of implementing a registration procedure based on classical
cross-correlation in Fourier Space, but of course you can use TurboReg
for instance.

Thomas

Emmanuele Sordini a écrit :

> Hello everybody,
> I am a new ImageJ user and ML member. My main interest in ImageJ stems
> from
> my amateur astronomy activities (astronomical imaging and image
> processing).
>
> Let's suppose the following scenario: I have a set of 16-bit B/W
> frames of
> the same target (typically, a planet or a deep-sky object) that have
> to be
> registered, sorted for quality and stacked. The biggest problem I'm faced
> with is what image quality indicator should be used for reliable sorting.
>
> The image quality metric should be easy to figure out and not very
> demanding
> in terms of computing power; yet, it should somehow reflect the relative
> abundance of details and features on the current frame, which (if I'm not
> mistaken) should result in relative abundance of high-frequency
> components
> in the FFT domain. The literature mentions quite a few "one-number"
> quantities: RMSE, SNR, peak SNR, and so on... but I'm quite puzzled. I've
> also come across some "perceptual indicators" which integrate
> quantitative
> measurement and quality as perceived by the human eye, but I'd rather not
> venture into this domain for the moment.
>
> Any ideas/suggestions/links/references on the subject will be highly
> valued.
>
> Thanks in advance,
> Emmanuele Sordini
>
> P.S. Should anybody be interested in my amateur astronomy work, please
> feel
> free to check out my website: www.bloomingstars.com
>
> .
>

--
/*****************************************************/
Thomas Boudier, MCU Université Paris 6,
Imagerie Integrative,Institut Curie - INSERM U759.
Tel : 01 69 86 31 72 Fax : 01 69 07 53 27
/****************************************************/

Emmanuele Sordini

Re: Image quality evaluation

Hello Thomas,

On 7/11/06, Thomas Boudier <[hidden email]> wrote:
>
>
> As concerns Astronomy i can suggest you to have a look at SalsaJ, a
> derivative from ImageJ that is customized for astronomy, it is
> multilingual and incorporates special features like aperture intensity
> measurement. You can have a look at :
> http://www.euhou.net/index.php?option=com_content&task=view&id=7&Itemid=9

Thanks for the link! I was not aware of the HOU initiative. I've just
downloaded the package and I will give it a try as soon as I can.

As about registration for astronomical images we are currently thinking
> of implementing a registration procedure based on classical
> cross-correlation in Fourier Space, but of course you can use TurboReg
> for instance.

Perhaps my first post was not very clear.

I am a long-time user of many software packages, but as such, I didn't care
much about the math background underlying many image processing features.
Now I thought about deepening my knowledge of the subject and dabbling in
some coding to see how the stuff "really" works. Here's why I went for
ImageJ: because it's a Java-based, open architecture, that allows people to
create and experiment with their own solutions.

For image registration, I first tried to use the Bij package: I must say it
works but it's also awfully slower (about an order of magnitude) than
Windows native applications. I will probably give TurboReg a try as well.

Do you have any suggestions as for the image quality indicator?

Thanks,
Emmanuele

Gabriel Landini

Re: Image quality evaluation

On Tuesday 11 July 2006 17:05, Emmanuele Sordini wrote:
> > measurement. You can have a look at :
> > http://www.euhou.net/index.php?option=com_content&task=view&id=7&Itemid=9
>

I noted that there is a webcam capture plugin (that is also supported in
Linux).
Does anybody know what webcams are supported and under what linux
distributions? (

Cheers,

Gabrile

Thomas Boudier-2

Re: Image quality evaluation

Hi Gabriel,

It is designed to work on modified version of a Philips webcam plugin
but it should work with normal ones, it uses the jmf (java media
framework http://java.sun.com/products/java-media/jmf/) api. I did not
tested it on linux but it should work on any distributions.

cheers

Thomas

Gabriel Landini a écrit :

> On Tuesday 11 July 2006 17:05, Emmanuele Sordini wrote:
>
>>> measurement. You can have a look at :
>>> http://www.euhou.net/index.php?option=com_content&task=view&id=7&Itemid=9
>>>
>
> I noted that there is a webcam capture plugin (that is also supported in
> Linux).
> Does anybody know what webcams are supported and under what linux
> distributions? (
>
> Cheers,
>
> Gabrile
>
> .
>
>

Jennifer West

Re: Image quality evaluation

In reply to this post by Emmanuele Sordini

Hi Emmanuele,

By chance I was working on this very problem yesterday!

I wrote a plugin that computes a "quality factor" based on a method
that I read at this site:
http://acquerra.com.au/astro/software/ppmcentre/

According to the webpage, this method is the same as the one used in
Registax.

I've only tested my plugin on two stacks of images (one of Jupiter
and one of Saturn) and it seems to work reasonably well. But I would
be very happy if you would like to test it and let me know how it
works on your images. I've just posted it to my website and you may
download it here:
http://www.umanitoba.ca/faculties/science/astronomy/jwest/plugins.html

My next step is to write a plugin that will automatically select the
images with the best quality and create a new stack.

Let me know how this works for you!

Jennifer

******************************************************
* Jennifer West
* Observatory Assistant/Outreach Coordinator
* Glenlea Astronomical Observatory, Campus Office
* 378 University College
* University of Manitoba
* Winnipeg, MB R3T 2M8
* P 204.474.9501
* F 204.261.0021
* www.umanitoba.ca/observatory
******************************************************

On 11-Jul-06, at 10:01 AM, Emmanuele Sordini wrote:

> Hello everybody,
> I am a new ImageJ user and ML member. My main interest in ImageJ
> stems from
> my amateur astronomy activities (astronomical imaging and image
> processing).
>
> Let's suppose the following scenario: I have a set of 16-bit B/W
> frames of
> the same target (typically, a planet or a deep-sky object) that
> have to be
> registered, sorted for quality and stacked. The biggest problem I'm
> faced
> with is what image quality indicator should be used for reliable
> sorting.
>
> The image quality metric should be easy to figure out and not very
> demanding
> in terms of computing power; yet, it should somehow reflect the
> relative
> abundance of details and features on the current frame, which (if
> I'm not
> mistaken) should result in relative abundance of high-frequency
> components
> in the FFT domain. The literature mentions quite a few "one-number"
> quantities: RMSE, SNR, peak SNR, and so on... but I'm quite
> puzzled. I've
> also come across some "perceptual indicators" which integrate
> quantitative
> measurement and quality as perceived by the human eye, but I'd
> rather not
> venture into this domain for the moment.
>
> Any ideas/suggestions/links/references on the subject will be
> highly valued.
>
> Thanks in advance,
> Emmanuele Sordini
>
> P.S. Should anybody be interested in my amateur astronomy work,
> please feel
> free to check out my website: www.bloomingstars.com

Breukers, C (TNW)

Re: Image quality evaluation

In reply to this post by Emmanuele Sordini

Hello Jennifer,

Processing and analyzing many images from fluorescent labeled cells I am
quite interested in determining a robust kind of quality for these
images.

Normally we use a Signal-to-Background Ratio equation:

((Mean of Max. intensity)-(Mean of Background intensity))/ Standard
Deviation of the background.

By applying this equation on our images we normally get ratios between
10 and 400. This is a direct factor between the mean signal and
background.

I have installed your plugin in ImageJ and it works fine, but I am
interested in the Results output file.

I get following result for a typical image: 1 2.156945272786E12

I am surprised by the large number. How should I interpret this number?

Regards,
Christian Breukers

------------------------------------
Biophysical Engineering Group
Faculty of Science and Technology
University of Twente

E-mail: [hidden email]
Web: www.tnw.utwente.nl/bpe
------------------------------------

Emmanuele Sordini

Re: Image quality evaluation

In reply to this post by Jennifer West

Hello Hennifer!

On 7/11/06, Jennifer West <[hidden email]> wrote:
>
> Hi Emmanuele,
>
>
> I wrote a plugin that computes a "quality factor" based on a method
> that I read at this site:
> http://acquerra.com.au/astro/software/ppmcentre/

According to the webpage, this method is the same as the one used in
> Registax.

Unfortunately, the source code of ppmcentre is not available (link to tar
file broken?), so I can't comment on that method. But if it's used in
Registax it should be quite reliable. Could you perhaps provide me with the
latest version (i.e. tar archive) you have?

BTW, my webcam processing application of choice is Iris. For sure it's way
less user-friendly than Registax, but it's also more powerful.

I've only tested my plugin on two stacks of images (one of Jupiter
> and one of Saturn) and it seems to work reasonably well. But I would
> be very happy if you would like to test it and let me know how it
> works on your images. I've just posted it to my website and you may
> download it here:
> http://www.umanitoba.ca/faculties/science/astronomy/jwest/plugins.html

Nice website! I also appreciate the "Save as FITS" plugin, which
inexplicably is not provided with ImageJ.

Let me know how this works for you!

I've done a very quick trial run on a 20-frame sample from an AVI file of
mine. Here's the outcome (frame numbers from 1 to 20 sorted in decreasing
quality order) as compared with the one from Iris:

Iris ImageQuality plugin
Frame # Score (times 10^11)
15 10 9.06731706255
19 15 9.06724426944
12 9 9.06694117833
17 1 9.06688137711
20 2 9.06688128297
10 18 9.06685968411
13 20 9.06681322122
11 6 9.06679849284
4 17 9.06679361922
8 7 9.06678465756
9 16 9.06666556767
18 19 9.06645265785
3 4 9.06644101764
5 12 9.06637309293
14 3 9.06632877069
1 8 9.06630080046
2 14 9.06627718458
7 13 9.06626186028
6 5 9.06622588776
16 11 9.06609830427

As you can see, it turned out to be quite different from Iris. But we know
that probably "no two image quality indicators are alike" :-)

Emmanuele

Emmanuele Sordini

Re: Image quality evaluation

In reply to this post by Breukers, C (TNW)

Hello Christian!

On 7/12/06, Breukers, C (TNW) <[hidden email]> wrote:
>
> Hello Jennifer,
>
> Normally we use a Signal-to-Background Ratio equation:
>
> ((Mean of Max. intensity)-(Mean of Background intensity))/ Standard
> Deviation of the background.

I would be interested in giving that indicator a try. However, I do have a
few questions:

1) How can you calculate the std. deviation of the background? i.e. how can
you tell the background from image subject (useful signal) and perform the
calculation only on the former? I'd rather take the std. deviation of the
whole image, which is of course close to that of the background if the
subject takes up only a fraction of the whole image.

2) For "Mean of max. intensity" and "mean of background intensity", see #1
as well. It all boils down to separating the signal and the background
correctly.

I would really appreciate it if you could elaborate a bit more on this
topic.

Emmanuele

dscho

Re: Image quality evaluation

Hi,

we used the entropy of the histogram for a long time, and then the entropy
after subtracting a massively blurred version of the image.

However, keep in mind that this (as the Extended Depth of Field plugin at
http://bigwww.epfl.ch/demo/edf/index.html) only measures the _crispness_
of the image.

So it does not actually work well in the presence of noise, since this
adds much more to the crispness than a well focused image. It should work
quite well to autofocus, though, since the amount of noise can be expected
to be fairly constant, no matter how focused the image is (or it should
actually add to the crispness _exactly_ in the case when the image is
focused).

Hth,
Dscho

Jennifer West

Re: Image quality evaluation

In reply to this post by Emmanuele Sordini

Hi Emmanuele and everyone,

Way back in July we were discussing image quality assessment. I have
been experimenting with using Sobel filters for assessing quality.
I've written a plugin that ranks images by their edges using the
"find edges" routine that is built into ImageJ.

If anyone is still working on this I would be interested to see how
the results from this routine compares to routines that others may be
using. Please let me know.

The plugin may be downloaded from:
http://www.umanitoba.ca/faculties/science/astronomy/jwest/plugins.html

Cheers,
Jennifer

___________________________________
Jennifer West, Instructor
Department of Physics and Astronomy
378 University College
University of Manitoba
Ph. 204-474-9501

On 12-Jul-06, at 4:06 AM, Emmanuele Sordini wrote:

> Hello Hennifer!
>
> On 7/11/06, Jennifer West <[hidden email]> wrote:
>>
>> Hi Emmanuele,
>>
>>
>> I wrote a plugin that computes a "quality factor" based on a method
>> that I read at this site:
>> http://acquerra.com.au/astro/software/ppmcentre/
>
> According to the webpage, this method is the same as the one used in
>> Registax.
>
>
> Unfortunately, the source code of ppmcentre is not available (link
> to tar
> file broken?), so I can't comment on that method. But if it's used in
> Registax it should be quite reliable. Could you perhaps provide me
> with the
> latest version (i.e. tar archive) you have?
>
> BTW, my webcam processing application of choice is Iris. For sure
> it's way
> less user-friendly than Registax, but it's also more powerful.
>
>
> I've only tested my plugin on two stacks of images (one of Jupiter
>> and one of Saturn) and it seems to work reasonably well. But I would
>> be very happy if you would like to test it and let me know how it
>> works on your images. I've just posted it to my website and you may
>> download it here:
>> http://www.umanitoba.ca/faculties/science/astronomy/jwest/
>> plugins.html
>
>
> Nice website! I also appreciate the "Save as FITS" plugin, which
> inexplicably is not provided with ImageJ.
>
> Let me know how this works for you!
>
>
> I've done a very quick trial run on a 20-frame sample from an AVI
> file of
> mine. Here's the outcome (frame numbers from 1 to 20 sorted in
> decreasing
> quality order) as compared with the one from Iris:
>
> Iris ImageQuality plugin
> Frame # Score (times 10^11)
> 15 10 9.06731706255
> 19 15 9.06724426944
> 12 9 9.06694117833
> 17 1 9.06688137711
> 20 2 9.06688128297
> 10 18 9.06685968411
> 13 20 9.06681322122
> 11 6 9.06679849284
> 4 17 9.06679361922
> 8 7 9.06678465756
> 9 16 9.06666556767
> 18 19 9.06645265785
> 3 4 9.06644101764
> 5 12 9.06637309293
> 14 3 9.06632877069
> 1 8 9.06630080046
> 2 14 9.06627718458
> 7 13 9.06626186028
> 6 5 9.06622588776
> 16 11 9.06609830427
>
> As you can see, it turned out to be quite different from Iris. But
> we know
> that probably "no two image quality indicators are alike" :-)
>
> Emmanuele

Nas-5

Re: Image quality evaluation

Hi Emmanuele,

I tried the link for Image Quality Calculator, but
there was no such plugin in there I'm afraid.

best,
Nas

--- Jennifer West <[hidden email]> wrote:

> Hi Emmanuele and everyone,
>
> Way back in July we were discussing image quality
> assessment. I have
> been experimenting with using Sobel filters for
> assessing quality.
> I've written a plugin that ranks images by their
> edges using the
> "find edges" routine that is built into ImageJ.
>
> If anyone is still working on this I would be
> interested to see how
> the results from this routine compares to routines
> that others may be
> using. Please let me know.
>
> The plugin may be downloaded from:
>

http://www.umanitoba.ca/faculties/science/astronomy/jwest/plugins.html

>
> Cheers,
> Jennifer
>
> ___________________________________
> Jennifer West, Instructor
> Department of Physics and Astronomy
> 378 University College
> University of Manitoba
> Ph. 204-474-9501
>
>
> On 12-Jul-06, at 4:06 AM, Emmanuele Sordini wrote:
>
> > Hello Hennifer!
> >
> > On 7/11/06, Jennifer West <[hidden email]>
> wrote:
> >>
> >> Hi Emmanuele,
> >>
> >>
> >> I wrote a plugin that computes a "quality factor"
> based on a method
> >> that I read at this site:
> >> http://acquerra.com.au/astro/software/ppmcentre/
> >
> > According to the webpage, this method is the same
> as the one used in
> >> Registax.
> >
> >
> > Unfortunately, the source code of ppmcentre is not
> available (link
> > to tar
> > file broken?), so I can't comment on that method.
> But if it's used in
> > Registax it should be quite reliable. Could you
> perhaps provide me
> > with the
> > latest version (i.e. tar archive) you have?
> >
> > BTW, my webcam processing application of choice is
> Iris. For sure
> > it's way
> > less user-friendly than Registax, but it's also
> more powerful.
> >
> >
> > I've only tested my plugin on two stacks of images
> (one of Jupiter
> >> and one of Saturn) and it seems to work
> reasonably well. But I would
> >> be very happy if you would like to test it and
> let me know how it
> >> works on your images. I've just posted it to my
> website and you may
> >> download it here:
> >>
>

http://www.umanitoba.ca/faculties/science/astronomy/jwest/

>
> >> plugins.html
> >
> >
> > Nice website! I also appreciate the "Save as FITS"
> plugin, which
> > inexplicably is not provided with ImageJ.
> >
> > Let me know how this works for you!
> >
> >
> > I've done a very quick trial run on a 20-frame
> sample from an AVI
> > file of
> > mine. Here's the outcome (frame numbers from 1 to
> 20 sorted in
> > decreasing
> > quality order) as compared with the one from Iris:
> >
> > Iris ImageQuality
> plugin
> > Frame # Score (times 10^11)
> > 15 10 9.06731706255
> > 19 15 9.06724426944
> > 12 9 9.06694117833
> > 17 1 9.06688137711
> > 20 2 9.06688128297
> > 10 18 9.06685968411
> > 13 20 9.06681322122
> > 11 6 9.06679849284
> > 4 17 9.06679361922
> > 8 7 9.06678465756
> > 9 16 9.06666556767
> > 18 19 9.06645265785
> > 3 4 9.06644101764
> > 5 12 9.06637309293
> > 14 3 9.06632877069
> > 1 8 9.06630080046
> > 2 14 9.06627718458
> > 7 13 9.06626186028
> > 6 5 9.06622588776
> > 16 11 9.06609830427
> >
> > As you can see, it turned out to be quite
> different from Iris. But
> > we know
> > that probably "no two image quality indicators are
> alike" :-)
> >
> > Emmanuele
>

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Aryeh Weiss

file rename

In reply to this post by Jennifer West

I am trying to use the following macro function:

> *File.rename(path1, path2)* - Renames, or moves, a file or directory.
> Returns "1" (true) if successful. The file or directory must be in the
> user's home directory, the ImageJ directory or the temp directory.
> Requires 1.38b.

I had previously used getDirectory to select a directory.
Must I specify the whole path for rename, or can it work in the
directory I have selected with
getDirectory? That directory is not the ImageJ directory, and not the
temp directory, and I dont think it is the home directory,
but there must be something obvious that I am missing.

--aryeh

--
Aryeh Weiss
School of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph: 972-3-5317638
FAX: 972-3-7384050