Convex Hull BG, a new (?) background subtraction algorithm

[Koen van dijken]

Hi fellow ImageJ users,

Last week I've been programming a background subtracter which is based on
another algorithm which I devised myself. I have not heard of it before. It
is based on building a 3D Convex Hull over the image's intensity profile.

I created a document about the motivation for building it, its workings,
its characteristics, how to use etc. Because the listserver will not let me
attach pdf's, here is a link to it on my dropbox:

https://www.dropbox.com/s/rubady3rv9yyxdv/Convex%20Hull%20BG.pdf?dl=0

Because gmail will not let me attach jars, here is a link to it on my
dropbox:

https://www.dropbox.com/s/1pxbney44su19e6/Convex_Hull_BG-0.9.jar?dl=0

Have fun trying it, and if you find any issues with it, please mail.

Koen van Dijken
Groningen, The Netherlands

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

[Michael Schmid]

Hi Koen,

the convex hull is a good background subtraction algorithm if the
background is convex, which is usually the case in brightfield light
microscopy. Unfortunately I could not try your plugin (I don't have Java
1.8).

I think that the concept is not entirely new, this paper describes a
similar approach (though different implementation):
   http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3433711/

You can do something that comes very close to it with ImageJ:
In 'Subtract Background', select 'Sliding Paraboloid' and a very large
radius (1000000 does not hurt). This gives a reasonably good
approximation of the convex hull. (It is an approximation not only
because the radius is finite but more so because it simplifies the
problem from 2D to several passes of 1D lines across the image, for
increased computing speed).

For color images you should have 'Separate Colors' enabled, then you
don't need to do the detour via a composite image.

Michael
________________________________________________________________
On 2016-07-17 16:10, Koen van dijken wrote:

> Hi fellow ImageJ users,
>
> Last week I've been programming a background subtracter which is based on
> another algorithm which I devised myself. I have not heard of it before. It
> is based on building a 3D Convex Hull over the image's intensity profile.
>
> I created a document about the motivation for building it, its workings,
> its characteristics, how to use etc. Because the listserver will not let me
> attach pdf's, here is a link to it on my dropbox:
>
> https://www.dropbox.com/s/rubady3rv9yyxdv/Convex%20Hull%20BG.pdf?dl=0
>
> Because gmail will not let me attach jars, here is a link to it on my
> dropbox:
>
> https://www.dropbox.com/s/1pxbney44su19e6/Convex_Hull_BG-0.9.jar?dl=0
>
> Have fun trying it, and if you find any issues with it, please mail.
>
> Koen van Dijken
> Groningen, The Netherlands
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

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

[Koen van dijken]

I created a Github repository with the sources for the plugin. Besides the
sources there are some remnants of the Eclipse-project in the github
repository. This is because I am a new 'push'-user of github and still
finding my way.

The url for the repository is:

https://github.com/kvdijken/CHBG

Also the sources for JSI and QuickHull3D are in the repository.

Koen


On Mon, Jul 18, 2016 at 12:29 PM, Michael Schmid <[hidden email]>
wrote:

> Hi Koen,
>
> the convex hull is a good background subtraction algorithm if the
> background is convex, which is usually the case in brightfield light
> microscopy. Unfortunately I could not try your plugin (I don't have Java
> 1.8).
>
> I think that the concept is not entirely new, this paper describes a
> similar approach (though different implementation):
>   http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3433711/
>
> You can do something that comes very close to it with ImageJ:
> In 'Subtract Background', select 'Sliding Paraboloid' and a very large
> radius (1000000 does not hurt). This gives a reasonably good approximation
> of the convex hull. (It is an approximation not only because the radius is
> finite but more so because it simplifies the problem from 2D to several
> passes of 1D lines across the image, for increased computing speed).
>
> For color images you should have 'Separate Colors' enabled, then you don't
> need to do the detour via a composite image.
>
> Michael
> ________________________________________________________________
>
> On 2016-07-17 16:10, Koen van dijken wrote:
>
>> Hi fellow ImageJ users,
>>
>> Last week I've been programming a background subtracter which is based on
>> another algorithm which I devised myself. I have not heard of it before.
>> It
>> is based on building a 3D Convex Hull over the image's intensity profile.
>>
>> I created a document about the motivation for building it, its workings,
>> its characteristics, how to use etc. Because the listserver will not let
>> me
>> attach pdf's, here is a link to it on my dropbox:
>>
>> https://www.dropbox.com/s/rubady3rv9yyxdv/Convex%20Hull%20BG.pdf?dl=0
>>
>> Because gmail will not let me attach jars, here is a link to it on my
>> dropbox:
>>
>> https://www.dropbox.com/s/1pxbney44su19e6/Convex_Hull_BG-0.9.jar?dl=0
>>
>> Have fun trying it, and if you find any issues with it, please mail.
>>
>> Koen van Dijken
>> Groningen, The Netherlands
>>
>> --
>> 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

[Jan Eglinger]

Dear Koen,

that's a nice plugin. I suggest that you also create an update site [1]
for it, so it's easier for users to install it.

To facilitate this, I transformed your github code into a Maven project
[2]. This avoids packaging the dependencies of this plugin into a single
uber-jar, which can lead to possible version skews when being used
together with other plugins that depend on different versions of the
same dependencies. Maven helps to manage these dependency versions, and
makes installing your plugin easy:

     mvn -Dimagej.app.directory=C:\Path\to\Your\ImageJ.app

This will install the jar file and its dependencies into your existing
installation.
Additionally, if you use the updater to upload the plugin to an update
site, it will also know about the requires dependencies.


Best,
Jan

[1]: http://imagej.net/Update_Sites
[2]: https://github.com/kvdijken/CHBG/pull/1



On 19.07.2016 00:24, Koen van dijken wrote:

> I created a Github repository with the sources for the plugin. Besides the
> sources there are some remnants of the Eclipse-project in the github
> repository. This is because I am a new 'push'-user of github and still
> finding my way.
>
> The url for the repository is:
>
> https://github.com/kvdijken/CHBG
>
> Also the sources for JSI and QuickHull3D are in the repository.
>
> Koen
>
>
> On Mon, Jul 18, 2016 at 12:29 PM, Michael Schmid <[hidden email]>
> wrote:
>
>> Hi Koen,
>>
>> the convex hull is a good background subtraction algorithm if the
>> background is convex, which is usually the case in brightfield light
>> microscopy. Unfortunately I could not try your plugin (I don't have Java
>> 1.8).
>>
>> I think that the concept is not entirely new, this paper describes a
>> similar approach (though different implementation):
>>   http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3433711/
>>
>> You can do something that comes very close to it with ImageJ:
>> In 'Subtract Background', select 'Sliding Paraboloid' and a very large
>> radius (1000000 does not hurt). This gives a reasonably good approximation
>> of the convex hull. (It is an approximation not only because the radius is
>> finite but more so because it simplifies the problem from 2D to several
>> passes of 1D lines across the image, for increased computing speed).
>>
>> For color images you should have 'Separate Colors' enabled, then you don't
>> need to do the detour via a composite image.
>>
>> Michael
>> ________________________________________________________________
>>
>> On 2016-07-17 16:10, Koen van dijken wrote:
>>
>>> Hi fellow ImageJ users,
>>>
>>> Last week I've been programming a background subtracter which is based on
>>> another algorithm which I devised myself. I have not heard of it before.
>>> It
>>> is based on building a 3D Convex Hull over the image's intensity profile.
>>>
>>> I created a document about the motivation for building it, its workings,
>>> its characteristics, how to use etc. Because the listserver will not let
>>> me
>>> attach pdf's, here is a link to it on my dropbox:
>>>
>>> https://www.dropbox.com/s/rubady3rv9yyxdv/Convex%20Hull%20BG.pdf?dl=0
>>>
>>> Because gmail will not let me attach jars, here is a link to it on my
>>> dropbox:
>>>
>>> https://www.dropbox.com/s/1pxbney44su19e6/Convex_Hull_BG-0.9.jar?dl=0
>>>
>>> Have fun trying it, and if you find any issues with it, please mail.
>>>
>>> Koen van Dijken
>>> Groningen, The Netherlands
>>>
>>> --
>>> 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
>

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

[Koen van dijken]

Thank you Jan. I merged your PR on github
I know about update sites, but have not had time yet to delve into it
deeply. I will do that in short time.

Koen


On Tue, Jul 19, 2016 at 3:05 PM, Jan Eglinger <[hidden email]>
wrote:

> Dear Koen,
>
> that's a nice plugin. I suggest that you also create an update site [1]
> for it, so it's easier for users to install it.
>
> To facilitate this, I transformed your github code into a Maven project
> [2]. This avoids packaging the dependencies of this plugin into a single
> uber-jar, which can lead to possible version skews when being used together
> with other plugins that depend on different versions of the same
> dependencies. Maven helps to manage these dependency versions, and makes
> installing your plugin easy:
>
>     mvn -Dimagej.app.directory=C:\Path\to\Your\ImageJ.app
>
> This will install the jar file and its dependencies into your existing
> installation.
> Additionally, if you use the updater to upload the plugin to an update
> site, it will also know about the requires dependencies.
>
>
> Best,
> Jan
>
> [1]: http://imagej.net/Update_Sites
> [2]: https://github.com/kvdijken/CHBG/pull/1
>
>
>
>
> On 19.07.2016 00:24, Koen van dijken wrote:
>
>> I created a Github repository with the sources for the plugin. Besides the
>> sources there are some remnants of the Eclipse-project in the github
>> repository. This is because I am a new 'push'-user of github and still
>> finding my way.
>>
>> The url for the repository is:
>>
>> https://github.com/kvdijken/CHBG
>>
>> Also the sources for JSI and QuickHull3D are in the repository.
>>
>> Koen
>>
>>
>> On Mon, Jul 18, 2016 at 12:29 PM, Michael Schmid <[hidden email]
>> >
>> wrote:
>>
>> Hi Koen,
>>>
>>> the convex hull is a good background subtraction algorithm if the
>>> background is convex, which is usually the case in brightfield light
>>> microscopy. Unfortunately I could not try your plugin (I don't have Java
>>> 1.8).
>>>
>>> I think that the concept is not entirely new, this paper describes a
>>> similar approach (though different implementation):
>>>   http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3433711/
>>>
>>> You can do something that comes very close to it with ImageJ:
>>> In 'Subtract Background', select 'Sliding Paraboloid' and a very large
>>> radius (1000000 does not hurt). This gives a reasonably good
>>> approximation
>>> of the convex hull. (It is an approximation not only because the radius
>>> is
>>> finite but more so because it simplifies the problem from 2D to several
>>> passes of 1D lines across the image, for increased computing speed).
>>>
>>> For color images you should have 'Separate Colors' enabled, then you
>>> don't
>>> need to do the detour via a composite image.
>>>
>>> Michael
>>> ________________________________________________________________
>>>
>>> On 2016-07-17 16:10, Koen van dijken wrote:
>>>
>>> Hi fellow ImageJ users,
>>>>
>>>> Last week I've been programming a background subtracter which is based
>>>> on
>>>> another algorithm which I devised myself. I have not heard of it before.
>>>> It
>>>> is based on building a 3D Convex Hull over the image's intensity
>>>> profile.
>>>>
>>>> I created a document about the motivation for building it, its workings,
>>>> its characteristics, how to use etc. Because the listserver will not let
>>>> me
>>>> attach pdf's, here is a link to it on my dropbox:
>>>>
>>>> https://www.dropbox.com/s/rubady3rv9yyxdv/Convex%20Hull%20BG.pdf?dl=0
>>>>
>>>> Because gmail will not let me attach jars, here is a link to it on my
>>>> dropbox:
>>>>
>>>> https://www.dropbox.com/s/1pxbney44su19e6/Convex_Hull_BG-0.9.jar?dl=0
>>>>
>>>> Have fun trying it, and if you find any issues with it, please mail.
>>>>
>>>> Koen van Dijken
>>>> Groningen, The Netherlands
>>>>
>>>> --
>>>> 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
>>
>>

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