Sensitive Thresholding?

[Wilson R S (AT)]

I am trying to develop system whereby I can piece together in focus
segments from each colour image in a stack to a single image(like a
jigsaw puzzle if you will). The stack is of the same particle taken at
different Z values.

I have tried using the plugins "stack focuser" and "extended depth of
focus", however both result in a "fuzzy" image.

My initial idea was to use a thresholding method, however this seams to
pick out the out of focus sections first. I have tried a few different
thresholding methods, colour and greyscale (on green channel seems to be
the best), but nothing I get is useful.

 

Has anyone tried to create this sort of method before? Or is there a
plugin? Any help would be much appreciated.

I have available a sample stack of the sort of thing I am collecting, it
is only 5 images because of file attachment limits, generally I would be
using around 15 images.

 

Thanks In advance

 

Rhodri Wilson

F.A.T./HeSaS

University of Glamorgan

Wales

UK

[Przemko]

Wilson R S (AT) wrote:

> I am trying to develop system whereby I can piece together in focus
> segments from each colour image in a stack to a single image(like a
> jigsaw puzzle if you will). The stack is of the same particle taken at
> different Z values.
>
> I have tried using the plugins "stack focuser" and "extended depth of
> focus", however both result in a "fuzzy" image.
>
> My initial idea was to use a thresholding method, however this seams to
> pick out the out of focus sections first. I have tried a few different
> thresholding methods, colour and greyscale (on green channel seems to be
> the best), but nothing I get is useful.
>
>  

Hi!
I posted on the same topic but got no reply. Basically, extended depth
of focus plugin failed me completely. I tried two different variants of
the plugin and played with the parameter. The images I got would range
from bad, through poor to disastrous. I tried Photoshop CS4 (the
previous ones do not have it) extended depth of focus tool and got a
great image the first time, and without fiddling. So my question, as a
follow up to yours, is why the extended depth of focus does not work in
ImageJ? Look at my previous post for more detail

Cheers!
Przemko


--
Przemko Tylzanowski PhD
Dept. Musculoskeletal Sciences
LSD & Joints
ON1 box 813

Univeristy of Leuven
Herestraat 49
3000 Leuven
Belgium

phone: [phone]
fax  : [phone]

[Gabriel Landini]

On Monday 30 November 2009  11:46:47 Wilson R S (AT) wrote:
> I have tried using the plugins "stack focuser" and "extended depth of
> focus", however both result in a "fuzzy" image.

Those procedures work fine if the stack is properly registered (eg some
microscopes do not move vertically orthogonal to the stage and so the image
seems to move while going down the stack) and that illumination does not
change from frame to frame. Without correction, there is no chance of getting
a good result.

But more importantly, one has to consider that for semi transparent samples
(for instance in bright field, phase contrast), the problem of extending the
depth of focus does not have a single solution. For example if you embed *semi
transparent* objects in gelatine, take photos at various depths and then run
the algorithm. You are likely to find several "good focus" points at the same
x,y position at various z. If the objects were opaque, occlusion would prevent
this to occur.
 
> My initial idea was to use a thresholding method, however this seams to
> pick out the out of focus sections first. I have tried a few different
> thresholding methods, colour and greyscale (on green channel seems to be
> the best), but nothing I get is useful.

There is extensive literature on this and (the related problem) autofocus.
Reading what has been done so far and what works, will save you a lot of time.

Cheers,

G.

[Stephan Saalfeld]

Hi,

that's not my field really, but did you try to align the images first as
suggested in your previous post?

See, e,g,

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

included in Fiji or, if that fails (usually it does not) you could try
it manually, setting some control points in image a and corresponding
points in image b and then use `Plugins->Transform->Landmark
Correspondences' to register using these points.

An even easier way: `Plugins->Registration->Linear Stack Alignment with
SIFT', the ancestor of the previously mentioned registration plugin.

For that functionality, you need mpicbg_.jar

http://pacific.mpi-cbg.de/cgi-bin/gitweb.cgi?p=fiji.git;a=blob_plain;f=precompiled/mpicbg_.jar

which is included in Fiji but works for ImageJ too.  Only
Register_Virtual_Stack_Slices has too many dependencies across Fiji to
install it without serious headache ;)

I am telling this because my first Google-hit for `extended depth of
focus' got me to this page:

http://www.janrik.net/ptools/ExtendedFocusPano12/index.html

that clearly makes a point about the need of ideally perfect
registration

Best,
Stephan


On Mon, 2009-11-30 at 12:47 +0100, Przemko wrote:

> Wilson R S (AT) wrote:
> > I am trying to develop system whereby I can piece together in focus
> > segments from each colour image in a stack to a single image(like a
> > jigsaw puzzle if you will). The stack is of the same particle taken at
> > different Z values.
> >
> > I have tried using the plugins "stack focuser" and "extended depth of
> > focus", however both result in a "fuzzy" image.
> >
> > My initial idea was to use a thresholding method, however this seams to
> > pick out the out of focus sections first. I have tried a few different
> > thresholding methods, colour and greyscale (on green channel seems to be
> > the best), but nothing I get is useful.
> >
> >  
> Hi!
> I posted on the same topic but got no reply. Basically, extended depth
> of focus plugin failed me completely. I tried two different variants of
> the plugin and played with the parameter. The images I got would range
> from bad, through poor to disastrous. I tried Photoshop CS4 (the
> previous ones do not have it) extended depth of focus tool and got a
> great image the first time, and without fiddling. So my question, as a
> follow up to yours, is why the extended depth of focus does not work in
> ImageJ? Look at my previous post for more detail
>
> Cheers!
> Przemko
>
>