Duplications at overlaps with stitching plugins

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Duplications at overlaps with stitching plugins

L'assegnista
Dear all,
One of the main problems with stitching are the deformations at the
image border as they hamper a perfect overlap. In such cases the
stitching plugins generally shows a duplication of the labelled pixels
in the overlapping region of two images (independently on the fusion
method). This is really annoying, particularily when counting cells or
precesses.
I know that this probably depends on the not sufficiently good quality
of my lenses, but in the past I obtained better results with Vias 2.2
(http://research.mssm.edu/cnic/tools-vias.html) a powerful (altough
manual..) freeware stitching software that  has no longer been
developped. I compared the two softwares on the same stacks of images
and I noticed that  duplications were much less frequent in vias.

I hypothesized that this may depend from a specific integration option
that in Vias is called "replace" (I really don't know what means in
mathematical terms).
Does anibody have suggestions on how to reduce duplications at
overlapping regions? Using Vias is not an option as it is completely
manual and works only one channel at the time. However would it be
possible to implement the stitching plugins starting from Vias?

Thanks in advance for any help

Federico Luzzati


--
Federico Luzzati, PhD
Researcher
Dept. Life Sciences and Systems Biology
University of Turin
Via Accademia Albertina, 13
10123 Torino
Neuroscience Institute Cavalieri Ottolenghi (NICO)
Regione Gonzole, 10
10043 Orbassano (Torino)
ITALY
tel. +39-011670-4683/ -6631

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Duplications at overlaps with stitching plugins

dscho
Hi Federico,

On Fri, 14 Jun 2013, Federico Luzzati wrote:

> One of the main problems with stitching are the deformations at the
> image border as they hamper a perfect overlap. In such cases the
> stitching plugins generally shows a duplication of the labelled pixels
> in the overlapping region of two images (independently on the fusion
> method). This is really annoying, particularily when counting cells or
> precesses.
>
> I know that this probably depends on the not sufficiently good quality
> of my lenses, but in the past I obtained better results with Vias 2.2
> (http://research.mssm.edu/cnic/tools-vias.html) a powerful (altough
> manual..) freeware stitching software that  has no longer been
> developped. I compared the two softwares on the same stacks of images
> and I noticed that duplications were much less frequent in vias.
>
> I hypothesized that this may depend from a specific integration option
> that in Vias is called "replace" (I really don't know what means in
> mathematical terms).
>
> Does anibody have suggestions on how to reduce duplications at
> overlapping regions? Using Vias is not an option as it is completely
> manual and works only one channel at the time. However would it be
> possible to implement the stitching plugins starting from Vias?

If I am not mistaken, Fiji's Distortion Correction plugin should help you.
I have not used it myself yet, but what I heard was that it is pretty easy
to use: to calibrate it, throw a bunch of overlapping images obtained with
the same setup at the plugin.. It will then figure out what it thinks is
the most likely deformation model and allows you to compensate for it.

Please let me know if it works for you,
Johannes

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Duplications at overlaps with stitching plugins

L'assegnista
In reply to this post by L'assegnista
Hi Johannes
thank you for your answer.. the problem is that I forgot to mention
that I'm stiching tile scans of confocal stacks.
My goal is ultimately to improve my method of 3D reconstruction from
serial sections acquired with a confocal microscope. We published this
method in 2011
(http://www.frontiersin.org/neurogenesis/10.3389/fnins.2011.00070/abstract.
In the original method stitching was performed in VIAS and registration
in Reconstruct, but now i'm trying to implement it in fiji and use
stitching plugins and TrackEM2.

As far as I have understood the distrotion Correction is included in
TrackEM2, thus I was wondering if I can calculate the coordinates of
each stack using the stitching plugins, import stacks in trackEM2
according to such coordinates and than apply the lens correction.

Do you have any suggestion?

Thank you very much again

Federico Luzzati

Il 14.06.2013 19:11 Johannes Schindelin ha scritto:

> Hi Federico,
>
> On Fri, 14 Jun 2013, Federico Luzzati wrote:
>
>> One of the main problems with stitching are the deformations at the
>> image border as they hamper a perfect overlap. In such cases the
>> stitching plugins generally shows a duplication of the labelled
>> pixels
>> in the overlapping region of two images (independently on the fusion
>> method). This is really annoying, particularily when counting cells
>> or
>> precesses.
>>
>> I know that this probably depends on the not sufficiently good
>> quality
>> of my lenses, but in the past I obtained better results with Vias
>> 2.2
>> (http://research.mssm.edu/cnic/tools-vias.html) a powerful (altough
>> manual..) freeware stitching software that  has no longer been
>> developped. I compared the two softwares on the same stacks of
>> images
>> and I noticed that duplications were much less frequent in vias.
>>
>> I hypothesized that this may depend from a specific integration
>> option
>> that in Vias is called "replace" (I really don't know what means in
>> mathematical terms).
>>
>> Does anibody have suggestions on how to reduce duplications at
>> overlapping regions? Using Vias is not an option as it is completely
>> manual and works only one channel at the time. However would it be
>> possible to implement the stitching plugins starting from Vias?
>
> If I am not mistaken, Fiji's Distortion Correction plugin should help
> you.
> I have not used it myself yet, but what I heard was that it is pretty
> easy
> to use: to calibrate it, throw a bunch of overlapping images obtained
> with
> the same setup at the plugin.. It will then figure out what it thinks
> is
> the most likely deformation model and allows you to compensate for
> it.
>
> Please let me know if it works for you,
> Johannes

--
Federico Luzzati, PhD
Researcher
Dept. Life Sciences and Systems Biology
University of Turin
Via Accademia Albertina, 13
10123 Torino
Neuroscience Institute Cavalieri Ottolenghi (NICO)
Regione Gonzole, 10
10043 Orbassano (Torino)
ITALY
tel. +39-011670-4683/ -6631

--
Federico Luzzati, PhD
Researcher
Dept. Life Sciences and Systems Biology
University of Turin
Via Accademia Albertina, 13
10123 Torino
Neuroscience Institute Cavalieri Ottolenghi (NICO)
Regione Gonzole, 10
10043 Orbassano (Torino)
ITALY
tel. +39-011670-4683/ -6631

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Duplications at overlaps with stitching plugins

Philip Ershler
Hello Federico,
      In the approach you mention, will you have the facilities to correct for the Point Spread Function (PSF) distortion inherent in 3D reconstructions of confocal data?

Phil

On Jun 15, 2013, at 4:23 PM, Federico Luzzati <[hidden email]> wrote:

> Hi Johannes
> thank you for your answer.. the problem is that I forgot to mention that I'm stiching tile scans of confocal stacks.
> My goal is ultimately to improve my method of 3D reconstruction from serial sections acquired with a confocal microscope. We published this method in 2011 (http://www.frontiersin.org/neurogenesis/10.3389/fnins.2011.00070/abstract. In the original method stitching was performed in VIAS and registration in Reconstruct, but now i'm trying to implement it in fiji and use stitching plugins and TrackEM2.
>
> As far as I have understood the distrotion Correction is included in TrackEM2, thus I was wondering if I can calculate the coordinates of each stack using the stitching plugins, import stacks in trackEM2 according to such coordinates and than apply the lens correction.
>
> Do you have any suggestion?
>
> Thank you very much again
>
> Federico Luzzati
>
> Il 14.06.2013 19:11 Johannes Schindelin ha scritto:
>> Hi Federico,
>>
>> On Fri, 14 Jun 2013, Federico Luzzati wrote:
>>
>>> One of the main problems with stitching are the deformations at the
>>> image border as they hamper a perfect overlap. In such cases the
>>> stitching plugins generally shows a duplication of the labelled pixels
>>> in the overlapping region of two images (independently on the fusion
>>> method). This is really annoying, particularily when counting cells or
>>> precesses.
>>>
>>> I know that this probably depends on the not sufficiently good quality
>>> of my lenses, but in the past I obtained better results with Vias 2.2
>>> (http://research.mssm.edu/cnic/tools-vias.html) a powerful (altough
>>> manual..) freeware stitching software that  has no longer been
>>> developped. I compared the two softwares on the same stacks of images
>>> and I noticed that duplications were much less frequent in vias.
>>>
>>> I hypothesized that this may depend from a specific integration option
>>> that in Vias is called "replace" (I really don't know what means in
>>> mathematical terms).
>>>
>>> Does anibody have suggestions on how to reduce duplications at
>>> overlapping regions? Using Vias is not an option as it is completely
>>> manual and works only one channel at the time. However would it be
>>> possible to implement the stitching plugins starting from Vias?
>>
>> If I am not mistaken, Fiji's Distortion Correction plugin should help you.
>> I have not used it myself yet, but what I heard was that it is pretty easy
>> to use: to calibrate it, throw a bunch of overlapping images obtained with
>> the same setup at the plugin.. It will then figure out what it thinks is
>> the most likely deformation model and allows you to compensate for it.
>>
>> Please let me know if it works for you,
>> Johannes
>
> --
> Federico Luzzati, PhD
> Researcher
> Dept. Life Sciences and Systems Biology
> University of Turin
> Via Accademia Albertina, 13
> 10123 Torino
> Neuroscience Institute Cavalieri Ottolenghi (NICO)
> Regione Gonzole, 10
> 10043 Orbassano (Torino)
> ITALY
> tel. +39-011670-4683/ -6631
>
> --
> Federico Luzzati, PhD
> Researcher
> Dept. Life Sciences and Systems Biology
> University of Turin
> Via Accademia Albertina, 13
> 10123 Torino
> Neuroscience Institute Cavalieri Ottolenghi (NICO)
> Regione Gonzole, 10
> 10043 Orbassano (Torino)
> ITALY
> tel. +39-011670-4683/ -6631
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Duplications at overlaps with stitching plugins

L'assegnista
Hi philip,
I’ve never tried to deconvolute my stacks, and in our lab we do not
have any commercial software to do it.
However you could certainly try, in theory it should improve the
quality of images but be sure you really need it!

best regards

Federico

Il 16.06.2013 07:35 Philip Ershler ha scritto:

> Hello Federico,
>       In the approach you mention, will you have the facilities to
> correct for the Point Spread Function (PSF) distortion inherent in 3D
> reconstructions of confocal data?
>
> Phil
>
> On Jun 15, 2013, at 4:23 PM, Federico Luzzati
> <[hidden email]> wrote:
>
>> Hi Johannes
>> thank you for your answer.. the problem is that I forgot to mention
>> that I'm stiching tile scans of confocal stacks.
>> My goal is ultimately to improve my method of 3D reconstruction from
>> serial sections acquired with a confocal microscope. We published this
>> method in 2011
>> (http://www.frontiersin.org/neurogenesis/10.3389/fnins.2011.00070/abstract.
>> In the original method stitching was performed in VIAS and
>> registration in Reconstruct, but now i'm trying to implement it in
>> fiji and use stitching plugins and TrackEM2.
>>
>> As far as I have understood the distrotion Correction is included in
>> TrackEM2, thus I was wondering if I can calculate the coordinates of
>> each stack using the stitching plugins, import stacks in trackEM2
>> according to such coordinates and than apply the lens correction.
>>
>> Do you have any suggestion?
>>
>> Thank you very much again
>>
>> Federico Luzzati
>>
>> Il 14.06.2013 19:11 Johannes Schindelin ha scritto:
>>> Hi Federico,
>>>
>>> On Fri, 14 Jun 2013, Federico Luzzati wrote:
>>>
>>>> One of the main problems with stitching are the deformations at
>>>> the
>>>> image border as they hamper a perfect overlap. In such cases the
>>>> stitching plugins generally shows a duplication of the labelled
>>>> pixels
>>>> in the overlapping region of two images (independently on the
>>>> fusion
>>>> method). This is really annoying, particularily when counting
>>>> cells or
>>>> precesses.
>>>>
>>>> I know that this probably depends on the not sufficiently good
>>>> quality
>>>> of my lenses, but in the past I obtained better results with Vias
>>>> 2.2
>>>> (http://research.mssm.edu/cnic/tools-vias.html) a powerful
>>>> (altough
>>>> manual..) freeware stitching software that  has no longer been
>>>> developped. I compared the two softwares on the same stacks of
>>>> images
>>>> and I noticed that duplications were much less frequent in vias.
>>>>
>>>> I hypothesized that this may depend from a specific integration
>>>> option
>>>> that in Vias is called "replace" (I really don't know what means
>>>> in
>>>> mathematical terms).
>>>>
>>>> Does anibody have suggestions on how to reduce duplications at
>>>> overlapping regions? Using Vias is not an option as it is
>>>> completely
>>>> manual and works only one channel at the time. However would it be
>>>> possible to implement the stitching plugins starting from Vias?
>>>
>>> If I am not mistaken, Fiji's Distortion Correction plugin should
>>> help you.
>>> I have not used it myself yet, but what I heard was that it is
>>> pretty easy
>>> to use: to calibrate it, throw a bunch of overlapping images
>>> obtained with
>>> the same setup at the plugin.. It will then figure out what it
>>> thinks is
>>> the most likely deformation model and allows you to compensate for
>>> it.
>>>
>>> Please let me know if it works for you,
>>> Johannes
>>
>> --
>> Federico Luzzati, PhD
>> Researcher
>> Dept. Life Sciences and Systems Biology
>> University of Turin
>> Via Accademia Albertina, 13
>> 10123 Torino
>> Neuroscience Institute Cavalieri Ottolenghi (NICO)
>> Regione Gonzole, 10
>> 10043 Orbassano (Torino)
>> ITALY
>> tel. +39-011670-4683/ -6631
>>
>> --
>> Federico Luzzati, PhD
>> Researcher
>> Dept. Life Sciences and Systems Biology
>> University of Turin
>> Via Accademia Albertina, 13
>> 10123 Torino
>> Neuroscience Institute Cavalieri Ottolenghi (NICO)
>> Regione Gonzole, 10
>> 10043 Orbassano (Torino)
>> ITALY
>> tel. +39-011670-4683/ -6631
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
Federico Luzzati, PhD
Researcher
Dept. Life Sciences and Systems Biology
University of Turin
Via Accademia Albertina, 13
10123 Torino
Neuroscience Institute Cavalieri Ottolenghi (NICO)
Regione Gonzole, 10
10043 Orbassano (Torino)
ITALY
tel. +39-011670-4683/ -6631

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Duplications at overlaps with stitching plugins

Philip Ershler
Hi Federico,
    We have a group in my Institute that is doing 3D reconstructions of the structure of the T-Tubules in mammalian cardiac cells. They have even gone so far as to calculate  the actual PSF of the confocal system by scanning fluorescent beads embedded in agar. At this sort of magnification the unconvolved images are quite badly distorted by the PSF.

Best,  Phil

On Jun 17, 2013, at 1:53 AM, Federico Luzzati <[hidden email]> wrote:

> Hi philip,
> I’ve never tried to deconvolute my stacks, and in our lab we do not have any commercial software to do it.
> However you could certainly try, in theory it should improve the quality of images but be sure you really need it!
>
> best regards
>
> Federico
>
> Il 16.06.2013 07:35 Philip Ershler ha scritto:
>> Hello Federico,
>>      In the approach you mention, will you have the facilities to
>> correct for the Point Spread Function (PSF) distortion inherent in 3D
>> reconstructions of confocal data?
>>
>> Phil
>>
>> On Jun 15, 2013, at 4:23 PM, Federico Luzzati
>> <[hidden email]> wrote:
>>
>>> Hi Johannes
>>> thank you for your answer.. the problem is that I forgot to mention that I'm stiching tile scans of confocal stacks.
>>> My goal is ultimately to improve my method of 3D reconstruction from serial sections acquired with a confocal microscope. We published this method in 2011 (http://www.frontiersin.org/neurogenesis/10.3389/fnins.2011.00070/abstract. In the original method stitching was performed in VIAS and registration in Reconstruct, but now i'm trying to implement it in fiji and use stitching plugins and TrackEM2.
>>>
>>> As far as I have understood the distrotion Correction is included in TrackEM2, thus I was wondering if I can calculate the coordinates of each stack using the stitching plugins, import stacks in trackEM2 according to such coordinates and than apply the lens correction.
>>>
>>> Do you have any suggestion?
>>>
>>> Thank you very much again
>>>
>>> Federico Luzzati
>>>
>>> Il 14.06.2013 19:11 Johannes Schindelin ha scritto:
>>>> Hi Federico,
>>>>
>>>> On Fri, 14 Jun 2013, Federico Luzzati wrote:
>>>>
>>>>> One of the main problems with stitching are the deformations at the
>>>>> image border as they hamper a perfect overlap. In such cases the
>>>>> stitching plugins generally shows a duplication of the labelled pixels
>>>>> in the overlapping region of two images (independently on the fusion
>>>>> method). This is really annoying, particularily when counting cells or
>>>>> precesses.
>>>>>
>>>>> I know that this probably depends on the not sufficiently good quality
>>>>> of my lenses, but in the past I obtained better results with Vias 2.2
>>>>> (http://research.mssm.edu/cnic/tools-vias.html) a powerful (altough
>>>>> manual..) freeware stitching software that  has no longer been
>>>>> developped. I compared the two softwares on the same stacks of images
>>>>> and I noticed that duplications were much less frequent in vias.
>>>>>
>>>>> I hypothesized that this may depend from a specific integration option
>>>>> that in Vias is called "replace" (I really don't know what means in
>>>>> mathematical terms).
>>>>>
>>>>> Does anibody have suggestions on how to reduce duplications at
>>>>> overlapping regions? Using Vias is not an option as it is completely
>>>>> manual and works only one channel at the time. However would it be
>>>>> possible to implement the stitching plugins starting from Vias?
>>>>
>>>> If I am not mistaken, Fiji's Distortion Correction plugin should help you.
>>>> I have not used it myself yet, but what I heard was that it is pretty easy
>>>> to use: to calibrate it, throw a bunch of overlapping images obtained with
>>>> the same setup at the plugin.. It will then figure out what it thinks is
>>>> the most likely deformation model and allows you to compensate for it.
>>>>
>>>> Please let me know if it works for you,
>>>> Johannes
>>>
>>> --
>>> Federico Luzzati, PhD
>>> Researcher
>>> Dept. Life Sciences and Systems Biology
>>> University of Turin
>>> Via Accademia Albertina, 13
>>> 10123 Torino
>>> Neuroscience Institute Cavalieri Ottolenghi (NICO)
>>> Regione Gonzole, 10
>>> 10043 Orbassano (Torino)
>>> ITALY
>>> tel. +39-011670-4683/ -6631
>>>
>>> --
>>> Federico Luzzati, PhD
>>> Researcher
>>> Dept. Life Sciences and Systems Biology
>>> University of Turin
>>> Via Accademia Albertina, 13
>>> 10123 Torino
>>> Neuroscience Institute Cavalieri Ottolenghi (NICO)
>>> Regione Gonzole, 10
>>> 10043 Orbassano (Torino)
>>> ITALY
>>> tel. +39-011670-4683/ -6631
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> Federico Luzzati, PhD
> Researcher
> Dept. Life Sciences and Systems Biology
> University of Turin
> Via Accademia Albertina, 13
> 10123 Torino
> Neuroscience Institute Cavalieri Ottolenghi (NICO)
> Regione Gonzole, 10
> 10043 Orbassano (Torino)
> ITALY
> tel. +39-011670-4683/ -6631
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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